Insights on Innovation, R&D, and IP
Perspectives on patents, scientific research, emerging technologies, and the strategies shaping modern R&D
Executive Summary
In 2024, US patent infringement jury verdicts totaled $4.19 billion across 72 cases. Twelve individual verdicts exceeded $100million. The largest single award—$857 million in General Access Solutions v.Cellco Partnership (Verizon)—exceeded the annual R&D budget of many mid-market technology companies. In the first half of 2025 alone, total damages reached an additional $1.91 billion.
The consequences of incomplete patent intelligence are not abstract. In what has become one of the most instructive IP disputes in recent history, Masimo’s pulse oximetry patents triggered a US import ban on certain Apple Watch models, forcing Apple to disable its blood oxygen feature across an entire product line, halt domestic sales of affected models, invest in a hardware redesign, and ultimately face a $634 million jury verdict in November 2025. Apple—a company with one of the most sophisticated intellectual property organizations on earth—spent years in litigation over technology it might have designed around during development.
For organizations with fewer resources than Apple, the risk calculus is starker. A mid-size materials company, a university spinout, or a defense contractor developing next-generation battery technology cannot absorb a nine-figure verdict or a multi-year injunction. For these organizations, the patent landscape analysis conducted during the development phase is the primary risk mitigation mechanism. The quality of that analysis is not a matter of convenience. It is a matter of survival.
And yet, a growing number of R&D and IP teams are conducting that analysis using general-purpose AI tools—ChatGPT, Claude, Microsoft Co-Pilot—that were never designed for patent intelligence and are structurally incapable of delivering it.
This report presents the findings of a controlled comparison study in which identical patent landscape queries were submitted to four AI-powered tools: Cypris (a purpose-built R&D intelligence platform),ChatGPT (OpenAI), Claude (Anthropic), and Microsoft Co-Pilot. Two technology domains were tested: solid-state lithium-sulfur battery electrolytes using garnet-type LLZO ceramic materials (freedom-to-operate analysis), and bio-based polyamide synthesis from castor oil derivatives (competitive intelligence).
The results reveal a significant and structurally persistent gap. In Test 1, Cypris identified over 40 active US patents and published applications with granular FTO risk assessments. Claude identified 12. ChatGPT identified 7, several with fabricated attribution. Co-Pilot identified 4. Among the patents surfaced exclusively by Cypris were filings rated as “Very High” FTO risk that directly claim the technology architecture described in the query. In Test 2, Cypris cited over 100 individual patent filings with full attribution to substantiate its competitive landscape rankings. No general-purpose model cited a single patent number.
The most active sectors for patent enforcement—semiconductors, AI, biopharma, and advanced materials—are the same sectors where R&D teams are most likely to adopt AI tools for intelligence workflows. The findings of this report have direct implications for any organization using general-purpose AI to inform patent strategy, competitive intelligence, or R&D investment decisions.
1. Methodology
A controlled comparative evaluation was conducted on March 27, 2026. An identical patent landscape query was submitted verbatim to each platform under standardized testing conditions. No follow-up prompts, clarifications, or iterative refinements were permitted, ensuring that each platform was evaluated based solely on its initial response.
The outputs were preserved in their original form and evaluated against predefined criteria using publicly verifiable patent records.
1.1 Query
Identify all active US patents and published applications filed in the last 5 years related to solid-state lithium-sulfur battery electrolytes using garnet-type ceramic materials. For each, provide the assignee, filing date, key claims, and current legal status. Highlight any patents that could pose freedom-to-operate risks for a company developing a Li₇La₃Zr₂O₁₂(LLZO)-based composite electrolyte with a polymer interlayer.
1.2 Tools Evaluated
1.3 Evaluation Criteria
Each response was evaluated using a consistent six-part scoring framework: patent coverage, assignee accuracy, filing metadata completeness, depth of claim analysis, quality of FTO risk stratification, and the presence of actionable strategic guidance.
Patent numbers, assignees, filing information, and legal status were independently checked against publicly available USPTO and WIPO records. The evaluation focused on the completeness, accuracy, and practical utility of each platform’s output rather than writing quality or presentation.
2. Findings
2.1 Coverage Gap
The most significant finding is the scale of the coverage differential. Cypris identified over 40 active US patents and published applications spanning LLZO-polymer composite electrolytes, garnet interface modification, polymer interlayer architectures, lithium-sulfur specific filings, and adjacent ceramic composite patents. The results were organized by technology category with per-patent FTO risk ratings.
Claude identified 12 patents organized in a four-tier risk framework. Its analysis was structurally sound and correctly flagged the two highest-risk filings (Solid Energies US 11,967,678 and the LLZO nanofiber multilayer US 11,923,501). It also identified the University ofMaryland/ Wachsman portfolio as a concentration risk and noted the NASA SABERS portfolio as a licensing opportunity. However, it missed the majority of the landscape, including the entire Corning portfolio, GM's interlayer patents, theKorea Institute of Energy Research three-layer architecture, and the HonHai/SolidEdge lithium-sulfur specific filing.
ChatGPT identified 7 patents, but the quality of attribution was inconsistent. It listed assignees as "Likely DOE /national lab ecosystem" and "Likely startup / defense contractor cluster" for two filings—language that indicates the model was inferring rather than retrieving assignee data. In a freedom-to-operate context, an unverified assignee attribution is functionally equivalent to no attribution, as it cannot support a licensing inquiry or risk assessment.
Co-Pilot identified 4 US patents. Its output was the most limited in scope, missing the Solid Energies portfolio entirely, theUMD/ Wachsman portfolio, Gelion/ Johnson Matthey, NASA SABERS, and all Li-S specific LLZO filings.
2.2 Critical Patents Missed by Public Models
The following table presents patents identified exclusively by Cypris that were rated as High or Very High FTO risk for the proposed technology architecture. None were surfaced by any general-purpose model.
2.3 Patent Fencing: The Solid Energies Portfolio
Cypris identified a coordinated patent fencing strategy by Solid Energies, Inc. that no general-purpose model detected at scale. Solid Energies holds at least four granted US patents and one published application covering LLZO-polymer composite electrolytes across compositions(US-12463245-B2), gradient architectures (US-12283655-B2), electrode integration (US-12463249-B2), and manufacturing processes (US-20230035720-A1). Claude identified one Solid Energies patent (US 11,967,678) and correctly rated it as the highest-priority FTO concern but did not surface the broader portfolio. ChatGPT and Co-Pilot identified zero Solid Energies filings.
The practical significance is that a company relying on any individual patent hit would underestimate the scope of Solid Energies' IP position. The fencing strategy—covering the composition, the architecture, the electrode integration, and the manufacturing method—means that identifying a single design-around for one patent does not resolve the FTO exposure from the portfolio as a whole. This is the kind of strategic insight that requires seeing the full picture, which no general-purpose model delivered
2.4 Assignee Attribution Quality
ChatGPT's response included at least two instances of fabricated or unverifiable assignee attributions. For US 11,367,895 B1, the listed assignee was "Likely startup / defense contractor cluster." For US 2021/0202983 A1, the assignee was described as "Likely DOE / national lab ecosystem." In both cases, the model appears to have inferred the assignee from contextual patterns in its training data rather than retrieving the information from patent records.
In any operational IP workflow, assignee identity is foundational. It determines licensing strategy, litigation risk, and competitive positioning. A fabricated assignee is more dangerous than a missing one because it creates an illusion of completeness that discourages further investigation. An R&D team receiving this output might reasonably conclude that the landscape analysis is finished when it is not.
3. Structural Limitations of General-Purpose Models for Patent Intelligence
3.1 Training Data Is Not Patent Data
Large language models are trained on web-scraped text. Their knowledge of the patent record is derived from whatever fragments appeared in their training corpus: blog posts mentioning filings, news articles about litigation, snippets of Google Patents pages that were crawlable at the time of data collection. They do not have systematic, structured access to the USPTO database. They cannot query patent classification codes, parse claim language against a specific technology architecture, or verify whether a patent has been assigned, abandoned, or subjected to terminal disclaimer since their training data was collected.
This is not a limitation that improves with scale. A larger training corpus does not produce systematic patent coverage; it produces a larger but still arbitrary sampling of the patent record. The result is that general-purpose models will consistently surface well-known patents from heavily discussed assignees (QuantumScape, for example, appeared in most responses) while missing commercially significant filings from less publicly visible entities (Solid Energies, Korea Institute of EnergyResearch, Shenzhen Solid Advanced Materials).
3.2 The Web Is Closing to Model Scrapers
The data access problem is structural and worsening. As of mid-2025, Cloudflare reported that among the top 10,000 web domains, the majority now fully disallow AI crawlers such as GPTBot andClaudeBot via robots.txt. The trend has accelerated from partial restrictions to outright blocks, and the crawl-to-referral ratios reveal the underlying tension: OpenAI's crawlers access approximately1,700 pages for every referral they return to publishers; Anthropic's ratio exceeds 73,000 to 1.
Patent databases, scientific publishers, and IP analytics platforms are among the most restrictive content categories. A Duke University study in 2025 found that several categories of AI-related crawlers never request robots.txt files at all. The practical consequence is that the knowledge gap between what a general-purpose model "knows" about the patent landscape and what actually exists in the patent record is widening with each training cycle. A landscape query that a general-purpose model partially answered in 2023 may return less useful information in 2026.
3.3 General-Purpose Models Lack Ontological Frameworks for Patent Analysis
A freedom-to-operate analysis is not a summarization task. It requires understanding claim scope, prosecution history, continuation and divisional chains, assignee normalization (a single company may appear under multiple entity names across patent records), priority dates versus filing dates versus publication dates, and the relationship between dependent and independent claims. It requires mapping the specific technical features of a proposed product against independent claim language—not keyword matching.
General-purpose models do not have these frameworks. They pattern-match against training data and produce outputs that adopt the format and tone of patent analysis without the underlying data infrastructure. The format is correct. The confidence is high. The coverage is incomplete in ways that are not visible to the user.
4. Comparative Output Quality
The following table summarizes the qualitative characteristics of each tool's response across the dimensions most relevant to an operational IP workflow.
5. Implications for R&D and IP Organizations
5.1 The Confidence Problem
The central risk identified by this study is not that general-purpose models produce bad outputs—it is that they produce incomplete outputs with high confidence. Each model delivered its results in a professional format with structured analysis, risk ratings, and strategic recommendations. At no point did any model indicate the boundaries of its knowledge or flag that its results represented a fraction of the available patent record. A practitioner receiving one of these outputs would have no signal that the analysis was incomplete unless they independently validated it against a comprehensive datasource.
This creates an asymmetric risk profile: the better the format and tone of the output, the less likely the user is to question its completeness. In a corporate environment where AI outputs are increasingly treated as first-pass analysis, this dynamic incentivizes under-investigation at precisely the moment when thoroughness is most critical.
5.2 The Diversification Illusion
It might be assumed that running the same query through multiple general-purpose models provides validation through diversity of sources. This study suggests otherwise. While the four tools returned different subsets of patents, all operated under the same structural constraints: training data rather than live patent databases, web-scraped content rather than structured IP records, and general-purpose reasoning rather than patent-specific ontological frameworks. Running the same query through three constrained tools does not produce triangulation; it produces three partial views of the same incomplete picture.
5.3 The Appropriate Use Boundary
General-purpose language models are effective tools for a wide range of tasks: drafting communications, summarizing documents, generating code, and exploratory research. The finding of this study is not that these tools lack value but that their value boundary does not extend to decisions that carry existential commercial risk.
Patent landscape analysis, freedom-to-operate assessment, and competitive intelligence that informs R&D investment decisions fall outside that boundary. These are workflows where the completeness and verifiability of the underlying data are not merely desirable but are the primary determinant of whether the analysis has value. A patent landscape that captures 10% of the relevant filings, regardless of how well-formatted or confidently presented, is a liability rather than an asset.
6. Test 2: Competitive Intelligence — Bio-Based Polyamide Patent Landscape
To assess whether the findings from Test 1 were specific to a single technology domain or reflected a broader structural pattern, a second query was submitted to all four tools. This query shifted from freedom-to-operate analysis to competitive intelligence, asking each tool to identify the top 10organizations by patent filing volume in bio-based polyamide synthesis from castor oil derivatives over the past three years, with summaries of technical approach, co-assignee relationships, and portfolio trajectory.
6.1 Query
6.2 Summary of Results
6.3 Key Differentiators
Verifiability
The most consequential difference in Test 2 was the presence or absence of verifiable evidence. Cypris cited over 100 individual patent filings with full patent numbers, assignee names, and publication dates. Every claim about an organization’s technical focus, co-assignee relationships, and filing trajectory was anchored to specific documents that a practitioner could independently verify in USPTO, Espacenet, or WIPO PATENT SCOPE. No general-purpose model cited a single patent number. Claude produced the most structured and analytically useful output among the public models, with estimated filing ranges, product names, and strategic observations that were directionally plausible. However, without underlying patent citations, every claim in the response requires independent verification before it can inform a business decision. ChatGPT and Co-Pilot offered thinner profiles with no filing counts and no patent-level specificity.
Data Integrity
ChatGPT’s response contained a structural error that would mislead a practitioner: it listed CathayBiotech as organization #5 and then listed “Cathay Affiliate Cluster” as a separate organization at #9, effectively double-counting a single entity. It repeated this pattern with Toray at #4 and “Toray(Additional Programs)” at #10. In a competitive intelligence context where the ranking itself is the deliverable, this kind of error distorts the landscape and could lead to misallocation of competitive monitoring resources.
Organizations Missed
Cypris identified Kingfa Sci. & Tech. (8–10 filings with a differentiated furan diacid-based polyamide platform) and Zhejiang NHU (4–6 filings focused on continuous polymerization process technology)as emerging players that no general-purpose model surfaced. Both represent potential competitive threats or partnership opportunities that would be invisible to a team relying on public AI tools.Conversely, ChatGPT included organizations such as ANTA and Jiangsu Taiji that appear to be downstream users rather than significant patent filers in synthesis, suggesting the model was conflating commercial activity with IP activity.
Strategic Depth
Cypris’s cross-cutting observations identified a fundamental chemistry divergence in the landscape:European incumbents (Arkema, Evonik, EMS) rely on traditional castor oil pyrolysis to 11-aminoundecanoic acid or sebacic acid, while Chinese entrants (Cathay Biotech, Kingfa) are developing alternative bio-based routes through fermentation and furandicarboxylic acid chemistry.This represents a potential long-term disruption to the castor oil supply chain dependency thatWestern players have built their IP strategies around. Claude identified a similar theme at a higher level of abstraction. Neither ChatGPT nor Co-Pilot noted the divergence.
6.4 Test 2 Conclusion
Test 2 confirms that the coverage and verifiability gaps observed in Test 1 are not domain-specific.In a competitive intelligence context—where the deliverable is a ranked landscape of organizationalIP activity—the same structural limitations apply. General-purpose models can produce plausible-looking top-10 lists with reasonable organizational names, but they cannot anchor those lists to verifiable patent data, they cannot provide precise filing volumes, and they cannot identify emerging players whose patent activity is visible in structured databases but absent from the web-scraped content that general-purpose models rely on.
7. Conclusion
This comparative analysis, spanning two distinct technology domains and two distinct analytical workflows—freedom-to-operate assessment and competitive intelligence—demonstrates that the gap between purpose-built R&D intelligence platforms and general-purpose language models is not marginal, not domain-specific, and not transient. It is structural and consequential.
In Test 1 (LLZO garnet electrolytes for Li-S batteries), the purpose-built platform identified more than three times as many patents as the best-performing general-purpose model and ten times as many as the lowest-performing one. Among the patents identified exclusively by the purpose-built platform were filings rated as Very High FTO risk that directly claim the proposed technology architecture. InTest 2 (bio-based polyamide competitive landscape), the purpose-built platform cited over 100individual patent filings to substantiate its organizational rankings; no general-purpose model cited as ingle patent number.
The structural drivers of this gap—reliance on training data rather than live patent feeds, the accelerating closure of web content to AI scrapers, and the absence of patent-specific analytical frameworks—are not transient. They are inherent to the architecture of general-purpose models and will persist regardless of increases in model capability or training data volume.
For R&D and IP leaders, the practical implication is clear: general-purpose AI tools should be used for general-purpose tasks. Patent intelligence, competitive landscaping, and freedom-to-operate analysis require purpose-built systems with direct access to structured patent data, domain-specific analytical frameworks, and the ability to surface what a general-purpose model cannot—not because it chooses not to, but because it structurally cannot access the data.
The question for every organization making R&D investment decisions today is whether the tools informing those decisions have access to the evidence base those decisions require. This study suggests that for the majority of general-purpose AI tools currently in use, the answer is no.
Study Disclosure
This comparative evaluation was commissioned and published by Cypris. The testing methodology, prompts, evaluation criteria, and underlying outputs have been documented to support independent review and replication.
All platform outputs were preserved in their original form. Patent data and material factual claims were cross-checked against USPTO Patent Center and WIPO PATENTSCOPE records as of March 27, 2026. Cypris was one of the platforms evaluated and therefore has a commercial interest in the findings.
The Patent Intelligence Gap - A Comparative Analysis of Verticalized AI-Patent Tools vs. General-Purpose Language Models for R&D Decision-Making
All Blogs
Claude is a formidable reasoner, but unaided it answers patent and scientific questions from training data — and training data is not the patent record. The constraint is not intelligence; it is access. Without a live connection, Claude can overlook recent filings, misstate priority dates, or fabricate a patent number with complete confidence. The Model Context Protocol (MCP) closes that gap. It connects Claude to an authoritative source, so the model retrieves real records and reasons over them rather than reconstructing them from memory.
MCP is the open standard Anthropic introduced in late 2024, now supported across every major AI platform. Within the Claude ecosystem, Claude Desktop, Claude Code, and Claude Science each act as an MCP host that can call external connectors. This article sets out how those connectors work, how to connect patent and scientific data to Claude, and why the connector you choose determines the quality of the answer far more than the act of connecting.
How MCP works in Claude
An MCP host — Claude Desktop, Claude Code, or Claude Science — runs a client that discovers available connectors and translates a request into structured tool calls. The connector authenticates to the data source, formats the query, and returns structured records; Claude then reasons over them in the conversation. Connectors are configured in Claude's settings, not built from scratch, and MCP's security model rests on OAuth-scoped tokens and read-only access — the controls that make connecting external data defensible in an enterprise setting.
The effect is consequential. A plain-language question in Claude becomes a genuine query against a patent or scientific source, and the returned records are available for Claude to analyze, summarize, and cite with provenance.
What you can connect
A growing set of open-source MCP connectors expose public patent and scientific sources to Claude. Connectors exist for USPTO data through Patent Public Search and the Open Data Portal, for the EPO through the OPS API, and for Google Patents through third-party APIs, alongside academic connectors for arXiv and PubMed. Independent projects such as Patent Connector link Claude directly to official patent-office data across multiple jurisdictions.
These connectors solve access. They let Claude retrieve records from a named authority in natural language, eliminating the copy-paste workflow and the transcription errors a model makes when it reads patent data off a web page.
Access is the easy part
Connecting Claude to a dataset is now trivial. Reasoning over it is not. A point connector hands Claude an undifferentiated stream of records from a single source and delegates all interpretation to the model — and the evidence on context engineering is unambiguous: flooding a model with a large, unscoped set of records degrades accuracy rather than improving it.
Most open-source connectors also cover a single source. A complete R&D question spans the patent record and the scientific literature at once, so answering it through point connectors means running several and reconciling their output by hand. For an isolated lookup that is acceptable; for prior art, freedom-to-operate, or landscape work, it reinstates the very fragmentation MCP was meant to eliminate.
Point connector versus domain-oriented agent
The decisive distinction is between a connector that exposes a dataset and an agent built around a domain. A domain-oriented agent is shaped around a field's data, ontology, and workflows, so retrieval is scoped before it ever reaches Claude's context. Instead of returning everything a keyword matches, it surfaces the high-signal patents and papers that bear on the question. Access alone does not make Claude reason well about patents; the domain layer does.
This matters most in Claude Science, Claude's environment for analytical research. Claude Science reasons powerfully over technical material but carries none of the competitive and landscape context held in the patent and scientific record. A domain-oriented agent connected through MCP supplies precisely that signal, so an agent reasoning about a research problem can also judge whether it aligns with where the field is heading.
Connecting patent data to Claude in practice
Cypris exposes its intelligence layer to Claude through an MCP server, so the competitive and landscape context it maintains connects directly into Claude Desktop, Claude Code, or Claude Science. Rather than handing Claude a broad dataset, it applies a proprietary R&D ontology over a corpus of more than 500 million patents and scientific papers to scope retrieval to what a question actually requires.
Cypris Q, the platform's agentic layer, runs prior art, white space, freedom-to-operate, and regulatory workflows and returns cited output; Agentic Monitoring keeps a position current as new records publish. Cypris operates under enterprise API partnerships with OpenAI, Anthropic, and Google, with enterprise-grade security, and serves hundreds of enterprise customers across pharmaceuticals, chemicals, advanced materials, and other regulated industries.
FAQ
Can Claude search patents using MCP?
Claude can search patents using MCP when a patent connector is added through its settings, with Claude Desktop and Claude Code acting as MCP hosts. Claude calls the connector's search and retrieval tools and reasons over the returned records, which lets it work from real filings rather than training data.
How do I connect patent data to Claude?
You connect patent data to Claude by adding an MCP connector in Claude's settings, then letting Claude call that connector's tools during a conversation. The connector authenticates to a patent source and returns structured records, so a plain-language question becomes a real query rather than a recall from memory.
What is Claude Science and how does it use MCP?
Claude Science is Claude's environment for analytical research work, and it supports MCP connectors. Because it is strong at reasoning but does not carry patent and competitive landscape context, connecting a domain-oriented agent through MCP supplies that external signal to its analysis.
What is the difference between Claude Desktop and Claude Code for MCP?
Claude Desktop and Claude Code are both MCP hosts that can call connectors, differing mainly in setting: Claude Desktop is the general assistant environment, while Claude Code is oriented to engineering workflows. Either can connect to a patent or scientific data source through MCP.
Which open-source MCP connectors work with Claude?
Open-source MCP connectors for Claude include ones for USPTO Patent Public Search and the Open Data Portal, the EPO OPS API, Google Patents through third-party APIs, and academic sources such as arXiv and PubMed. Most cover a single source, so spanning patents and literature usually means running several.
Is connecting Claude to a dataset enough for patent research?
Connecting Claude to a dataset solves access but not reasoning, because a raw connector floods the model with records and an overwhelmed model reasons less accurately. Pairing retrieval with a domain ontology, so only high-signal records reach Claude, is what produces reliable analysis.
What is the difference between a point connector and a domain-oriented agent?
A point connector exposes one dataset and leaves interpretation to Claude, while a domain-oriented agent is built around a field's data, ontology, and workflows and scopes retrieval before it reaches the model. The connector improves retrieval; the agent improves the answer.
Can Cypris and Claude be used together?
Cypris and Claude can be used together, because Cypris exposes its intelligence layer through an MCP server and Claude supports MCP connectors, including in Claude Science. The landscape and competitive context Cypris maintains can be connected into Claude so an agent draws on external signal while it reasons.
Are MCP connectors secure for enterprise use with Claude?
MCP's security model relies on OAuth-scoped tokens and read-only access patterns, which is what makes connecting external data to Claude viable for enterprise use. Enterprise deployments should also confirm workspace-level controls and how data is handled with the underlying model provider.
What is the best way to give Claude patent and scientific data?
The best way to give Claude patent and scientific data for R&D work is a domain-oriented agent rather than a raw connector, because stage-gate work spans patents and literature and requires reasoning, not just retrieval. Cypris connects to Claude through an MCP server over a corpus of more than 500 million patents and scientific papers organized by a proprietary R&D ontology.
.jpg)
Regulatory intelligence is the discipline of tracking the approvals, submissions, guidance, and standards that decide whether a technology can reach the market. In regulated industries it stands alongside patent and scientific intelligence as a gate on every R&D program. A technology can be genuinely novel, fully patent-clear, and still be blocked, delayed, or reshaped by a single regulatory decision.
The signals are public but scattered across many bodies and formats: approvals and clearances, submission and trial records, guidance documents and rule changes, standards, labeling, and safety actions. Their value is highest early — before a rule change or a competitor's approval is widely understood. This article sets out how AI-powered regulatory intelligence works for R&D teams in 2026, and how it connects to the patent and scientific record.
What regulatory intelligence covers
Regulatory intelligence spans the full regulatory footprint of a technology area: approvals and clearances, submissions and clinical or field trial records, agency guidance and rule changes, technical standards, labeling requirements, and safety actions such as recalls. The relevant bodies differ by sector — drug and device regulators, environmental and chemical agencies, standards organizations — but the task is constant: know what has changed, what is pending, and what it means for a program.
The payoff is lead time and avoided risk. A competitor's submission reveals its direction and timeline. A guidance change can open or foreclose a development path. Catching either early is the difference between steering a program and being overtaken by a decision after the fact.
Why manual regulatory tracking lags
Manual regulatory tracking means monitoring dozens of agency websites and databases separately, then compiling findings by hand. It is slow, and it is partial. Keyword-based tracking misses documents that describe the same technology or requirement in different terms, and single-source monitoring severs the connection between a regulatory signal and the patent or scientific activity around the same technology.
It is also episodic. A periodic regulatory report is stale the moment a new decision publishes, and the window between refreshes is precisely where a missed signal becomes a missed deadline. Rising regulatory activity across sectors only widens that gap.
How AI-powered regulatory intelligence works
AI-powered regulatory intelligence replaces periodic keyword monitoring with continuous, meaning-based retrieval. Semantic search surfaces relevant approvals, submissions, and guidance by concept, so a signal registers even when it uses unfamiliar terminology. An R&D ontology organizes those signals by technology domain, tying each regulatory event to the specific technology and the organizations pursuing it.
Continuous monitoring runs the analysis without waiting for a scheduled review. It interprets each new regulatory signal against a defined domain, separates the material from the routine, and delivers contextualized alerts rather than raw document links. Because agents span sources, regulatory events can be correlated with patents, scientific literature, and corporate activity into a single picture of where a technology and its competitors are moving.
Connecting regulatory signals to patents and science
Regulatory intelligence is most valuable when it is not siloed. A regulatory decision is one input to a stage-gate, alongside prior art, freedom-to-operate, and the competitive landscape. Connecting regulatory signals to the patent and scientific record lets a team see that a competitor's approval aligns with a filing cluster and a research push — a far stronger signal than any one source read alone.
This is the shift AI enables: from monitoring agencies one at a time to interpreting regulatory change in the context of the full technology picture, and from a static report to intelligence that updates the moment decisions publish.
Regulatory intelligence in practice
Cypris is an AI-native R&D intelligence platform whose Agentic Monitoring capability tracks regulatory bodies continuously, alongside patent offices, scientific literature, M&A activity, product launches, grant awards, and corporate news. It interprets these signals through a proprietary R&D ontology over a corpus of more than 500 million patents and scientific papers, so a regulatory event is tied to the technology and the organizations it concerns rather than read in isolation.
Cypris Q, the platform's agentic layer, lets teams move from a regulatory signal into prior art, white space, or freedom-to-operate analysis on the same technology, in one environment, with cited output. Cypris operates under enterprise API partnerships with OpenAI, Anthropic, and Google, with enterprise-grade security, and serves hundreds of enterprise customers across pharmaceuticals, chemicals, advanced materials, and other regulated industries.
FAQ
What is regulatory intelligence for R&D?
Regulatory intelligence for R&D is the practice of tracking the approvals, submissions, guidance, and standards that determine whether a technology can reach the market. It sits alongside patent and scientific intelligence as a gate on a program, because a technology can be patent-clear and still be blocked or delayed by a regulatory decision.
How is regulatory intelligence different from patent monitoring?
Regulatory intelligence tracks regulatory signals such as approvals, submissions, and guidance, while patent monitoring tracks filings. Both gate an R&D program, and the fullest picture comes from correlating them, since a competitor's approval often aligns with its patent and research activity.
What signals does regulatory intelligence track?
Regulatory intelligence tracks approvals and clearances, submissions and trial records, agency guidance and rule changes, technical standards, labeling requirements, and safety actions such as recalls. The relevant bodies vary by sector, but the task is to know what has changed, what is pending, and what it means.
Why does regulatory intelligence matter for R&D?
Regulatory intelligence matters for R&D because a regulatory decision can open or close a development path regardless of a technology's novelty or patent position. Catching a guidance change or a competitor's submission early is the difference between adjusting a program and being caught by a decision after the fact.
How does AI improve regulatory intelligence?
AI improves regulatory intelligence by replacing periodic keyword monitoring with continuous semantic retrieval, so relevant approvals, submissions, and guidance are found by concept even when terminology differs. An R&D ontology then organizes the signals by domain and connects them to the technology and organizations involved.
Can regulatory signals be tracked continuously?
Regulatory signals can be tracked continuously with agentic monitoring that interprets new decisions against a defined technology domain and delivers contextualized alerts as they publish. This replaces periodic manual reports, which are stale as soon as a new decision appears.
How does regulatory intelligence connect to patents and science?
Regulatory intelligence connects to patents and science when the same platform correlates a regulatory event with the filings and research around the same technology. This produces a stronger signal than any single source, and it lets a regulatory decision feed directly into prior art or freedom-to-operate review.
Which sectors rely most on regulatory intelligence?
Regulated industries rely most on regulatory intelligence, including pharmaceuticals, medical devices, chemicals, advanced materials, and energy, where approvals and standards gate commercialization. In these sectors a regulatory signal can reshape an R&D program's timeline and direction.
What public sources support regulatory intelligence?
Public sources that support regulatory intelligence include agency databases and registers such as those published by drug, device, environmental, and standards bodies, along with trial registries and official rule-change publications. Unifying and interpreting these fragmented sources is what an AI-powered platform adds.
What is the best platform for regulatory intelligence in R&D?
The best platform for regulatory intelligence in R&D tracks regulatory signals continuously and connects them to the patent and scientific record. Cypris tracks regulatory bodies through Agentic Monitoring alongside patents, literature, and corporate signals, interpreted through a proprietary R&D ontology over a corpus of more than 500 million patents and scientific papers.
ChatGPT is the assistant many R&D and IP teams already use, but on its own it answers patent questions from training data. It can miss recent filings, confuse filing and publication dates, or produce a patent number that does not exist. Connecting ChatGPT to a live source through the Model Context Protocol (MCP) fixes this, so it retrieves real records and reasons over them.
MCP is an open standard introduced by Anthropic in late 2024 and now supported across the major AI platforms, ChatGPT among them. This article explains how ChatGPT's connectors and apps work, how to connect patent and scientific data, and why the choice of connector determines whether the output is reliable.
How connectors and apps work in ChatGPT
ChatGPT connects to external data through MCP-based apps. OpenAI renamed connectors to apps in December 2025, and in 2026 moved the app directory into a broader plugin directory, but the underlying mechanism is unchanged: an app is an MCP integration that lets ChatGPT call approved tools and retrieve information from a service. Custom MCP servers are added through Developer Mode, and on workspace plans administrators control whether custom apps are allowed and how they roll out.
Once connected, ChatGPT can call the app's tools during a chat or in deep research, so a plain-language question becomes a structured query against a patent or scientific source. MCP's security model relies on OAuth-scoped tokens and read-only access patterns, which keeps the connection appropriate for enterprise use.
What you can connect
Several open-source MCP servers expose public patent and scientific sources to ChatGPT. There are connectors for USPTO data through Patent Public Search and the Open Data Portal, for the EPO through the OPS API, and for Google Patents through third-party APIs, alongside academic connectors for arXiv and PubMed. Independent projects such as Patent Connector link ChatGPT directly to official patent-office data across several jurisdictions.
These connectors solve access. They let ChatGPT retrieve records from a specific authority in natural language, which removes the manual copy-paste loop and the errors a model makes when it reads patent data off a web page.
Access is the easy part
Connecting ChatGPT to a dataset is now straightforward. Reasoning over it well is the harder problem. A point connector hands ChatGPT a stream of raw records from one source and leaves interpretation to the model, and research on context engineering shows that flooding a model with a large, undifferentiated set of records degrades accuracy rather than improving it.
Most open-source connectors also cover a single source, so a question that spans the patent record and the scientific literature usually means running several apps and reconciling their output by hand. That is acceptable for a quick lookup but not for prior art, freedom-to-operate, or landscape work.
Point connector versus domain-oriented agent
The meaningful distinction is between an app that exposes a dataset and an agent built around a domain. A domain-oriented agent is shaped around a field's data, ontology, and workflows, so retrieval is scoped before it reaches ChatGPT's context. Rather than returning everything a keyword matches, it retrieves the high-signal patents and papers relevant to the question. Access alone does not make ChatGPT reason well about patents; the domain layer does.
For teams on workspace plans, this also simplifies governance. A single domain-oriented app under administrator control is easier to manage and audit than a stack of point connectors, each with its own source, credentials, and maintenance burden.
Connecting patent data to ChatGPT in practice
Cypris exposes its intelligence layer through an MCP server, so its competitive and landscape context can be connected into ChatGPT as an app. Rather than handing ChatGPT a broad dataset, it uses a proprietary R&D ontology over a corpus of more than 500 million patents and scientific papers to scope retrieval to what matters for a question, and returns source-traceable results ChatGPT can cite.
Cypris Q, the platform's agentic layer, runs prior art, white space, freedom-to-operate, and regulatory workflows and returns cited output, and Agentic Monitoring keeps a position current as new records publish. Cypris operates under enterprise API partnerships with OpenAI, Anthropic, and Google, with enterprise-grade security, and serves hundreds of enterprise customers across pharmaceuticals, chemicals, advanced materials, and other regulated industries.
FAQ
Can ChatGPT search patents using MCP?
ChatGPT can search patents using MCP when a patent app or connector is added, after which it calls the app's tools to retrieve records during a chat or deep research. This lets ChatGPT work from real filings rather than training data, which removes the hallucination and stale-coverage problems of answering from memory.
How do I connect patent data to ChatGPT?
You connect patent data to ChatGPT by adding an MCP-based app, typically a custom MCP server through Developer Mode, subject to any workspace controls. Once connected, ChatGPT can call the app's tools so a plain-language question becomes a structured query against a patent source.
What are ChatGPT apps and connectors?
ChatGPT apps are MCP integrations that let ChatGPT call approved tools and retrieve information from a service; OpenAI renamed connectors to apps in December 2025 and later organized them in a plugin directory. The mechanism is MCP, so the same standard used by other assistants applies.
What is Developer Mode in ChatGPT?
Developer Mode is the setting that lets you add custom MCP servers to ChatGPT beyond the built-in apps. It is how a team connects a specific patent or scientific data source that is not already offered as a packaged app.
Which open-source MCP connectors work with ChatGPT?
Open-source MCP connectors for ChatGPT include ones for USPTO Patent Public Search and the Open Data Portal, the EPO OPS API, Google Patents through third-party APIs, and academic sources such as arXiv and PubMed. Most cover a single source, so spanning patents and literature usually means running several.
Is connecting ChatGPT to a dataset enough for patent research?
Connecting ChatGPT to a dataset solves access but not reasoning, because a raw connector floods the model with records and an overwhelmed model reasons less accurately. Pairing retrieval with a domain ontology, so only high-signal records reach ChatGPT, is what produces reliable analysis.
How do enterprise controls work for ChatGPT apps?
On workspace plans, administrators control whether custom apps are allowed and how they roll out, which lets an organization govern what data ChatGPT can reach. Combined with MCP's OAuth-scoped, read-only access model, this is what makes connecting external data appropriate for enterprise use.
What is the difference between a point connector and a domain-oriented agent?
A point connector exposes one dataset and leaves interpretation to ChatGPT, while a domain-oriented agent is built around a field's data, ontology, and workflows and scopes retrieval before it reaches the model. The connector improves retrieval; the agent improves the answer, and it is also easier to govern as a single app.
Can Cypris and ChatGPT be used together?
Cypris and ChatGPT can be used together, because Cypris exposes its intelligence layer through an MCP server and ChatGPT connects to MCP servers as apps. The landscape and competitive context Cypris maintains can be connected into ChatGPT so it reasons over scoped, source-traceable records.
What is the best way to give ChatGPT patent and scientific data?
The best way to give ChatGPT patent and scientific data for R&D work is a domain-oriented agent rather than a raw connector, because stage-gate work spans patents and literature and requires reasoning, not just retrieval. Cypris connects to ChatGPT through an MCP server over a corpus of more than 500 million patents and scientific papers organized by a proprietary R&D ontology.
Keyword search matches exact terms. Semantic search matches meaning. For patent search, that distinction determines whether a strategically critical filing is found or missed.
Patent search has relied on Boolean keyword queries and classification codes for decades. The method works when the searcher already knows the exact language an invention will use. It fails when a competitor describes the same mechanism with different words, files under a different classification, or uses terminology that did not exist when the query was written. In fast-moving fields, that failure is routine.
In 2026, R&D and IP teams are moving to AI-native semantic patent search because the volume and linguistic variety of global filings have outpaced keyword methods. This article defines semantic search, contrasts it with keyword search, and explains what the shift changes for patent search, patent analytics, prior art, and freedom-to-operate work.
How keyword patent search works and where it breaks
Keyword search retrieves documents that contain the specific terms in a query, usually combined with Boolean operators and classification filters. It is precise when the vocabulary is known and stable, and it remains useful for targeted lookups.
It breaks on vocabulary mismatch. Two teams working on the same problem often use entirely different terminology, and patent drafters frequently choose broad or unusual language deliberately. A keyword query built around expected terms will not retrieve a filing that describes the same invention differently. The result is silent gaps: the searcher sees results and assumes coverage, without knowing what was missed.
Volume magnifies the problem. Global patent filings and scientific publications continue to rise, and the World Intellectual Property Organization reported scientific output above two million articles in 2025. Expanding keyword queries to chase this volume produces either too much noise or too little signal.
How semantic search works
Semantic search represents the meaning of text as mathematical vectors, so that conceptually similar passages sit close together regardless of exact wording. A query for a mechanism retrieves filings that describe that mechanism, even when the words differ. This directly addresses the vocabulary-mismatch problem that keyword search cannot solve.
For patents, the strongest implementations apply semantic search at the claim level and across both patents and scientific literature. Claim-level retrieval matters because the legal risk in a patent lives in its claims, not its abstract. Searching patents and scientific papers together matters because early technical disclosure often appears in the literature before it reaches granted claims.
An R&D ontology strengthens semantic search further. An ontology is a structured map of technical concepts and their relationships. When semantic retrieval is organized through an ontology, as it is on AI-native platforms such as Cypris, the system interprets a query in the context of a technology domain rather than as isolated words, which improves both recall and precision.
What the shift changes for R&D and IP teams
Semantic search changes prior art and FTO work most directly. In prior art search, semantic retrieval surfaces conceptually relevant disclosures that keyword queries overlook, which strengthens both patentability assessments and invalidity arguments. In freedom-to-operate search, it surfaces active claims a product may read on even when those claims use unexpected language, reducing unquantified legal risk.
It also changes patent analytics. Once retrieval understands meaning, analytics can group filings by technical concept rather than by literal text, producing cleaner technology landscapes, competitor maps, and white space analysis. Agentic workflows build on this by chaining retrieval and reasoning steps to assemble landscapes, comparison matrices, and monitored positions automatically.
Semantic search in practice
Cypris is an AI-native R&D intelligence platform built on semantic search across a corpus of more than 500 million patents and scientific papers, organized through a proprietary R&D ontology. The ontology lets Cypris interpret technical meaning and retrieve conceptually related patents and literature at the claim level, rather than matching keywords.
Cypris Q, the platform's agentic layer, chains semantic retrieval and reasoning into end-to-end workflows such as landscape analysis, prior art review, and FTO assessment. Agentic Monitoring keeps those positions current by evaluating new filings as they publish. Cypris operates under enterprise API partnerships with OpenAI, Anthropic, and Google, with enterprise-grade security, and serves hundreds of enterprise customers across pharmaceuticals, chemicals, advanced materials, and other regulated industries.
FAQ
What is semantic search for patents?
Semantic search for patents retrieves filings by meaning rather than by exact keywords, representing text as vectors so that conceptually similar patents sit close together. This surfaces relevant patents that use different terminology than a query expects, which keyword search cannot do.
What is the difference between semantic search and keyword search?
Semantic search matches the meaning of text, while keyword search matches exact terms combined with Boolean operators. Keyword search misses filings that describe the same invention in different words, whereas semantic search retrieves them because it operates on concepts rather than literal strings.
Why are R&D teams moving to AI-native patent search?
R&D teams are moving to AI-native patent search because the volume and linguistic variety of global filings have outpaced keyword methods, causing silent gaps in coverage. Semantic search retrieves conceptually related filings across patents and scientific literature, reducing the risk that critical disclosures are missed.
Is semantic search better than keyword search for prior art?
Semantic search is generally stronger for prior art because it surfaces conceptually relevant disclosures that keyword queries overlook due to vocabulary mismatch. Keyword search remains useful for targeted lookups when the exact terminology is known, so many workflows combine both.
What is an R&D ontology in patent search?
An R&D ontology is a structured map of technical concepts and their relationships that organizes a search corpus by meaning. In patent search, an ontology lets a system interpret a query in the context of a technology domain rather than as isolated words, improving both recall and precision.
Does semantic search work across patents and scientific papers?
Semantic search works across both patents and scientific papers when the corpus unifies them, which matters because early technical disclosure often appears in the literature before it reaches granted patent claims. Searching both together produces a more complete technical and competitive picture.
How does semantic search improve patent analytics?
Semantic search improves patent analytics by grouping filings by technical concept rather than literal text, which produces cleaner technology landscapes, competitor maps, and white space analysis. Analytics built on meaning are more reliable than analytics built on keyword matches alone.
Can semantic patent search be automated with agents?
Semantic patent search can be automated with agentic workflows that chain retrieval and reasoning steps to assemble landscapes, comparison matrices, and monitored positions. Agents keep the analysis current by re-running semantic retrieval against new filings as they publish.
Does semantic search replace Boolean patent search entirely?
Semantic search does not fully replace Boolean patent search, because targeted keyword queries remain useful when exact terminology is known. The strongest workflows combine semantic retrieval for recall with keyword precision for confirmation.
What data coverage does effective semantic patent search require?
Effective semantic patent search requires broad coverage across patents and scientific literature, so that conceptually related disclosures in any vocabulary can be retrieved. A corpus of more than 500 million patents and scientific papers organized through an R&D ontology supports this breadth.
R&D knowledge management is the practice of capturing, organizing, and making retrievable the knowledge a research organization generates, so it accumulates instead of dissipating. Every program produces reports, experiments, analyses, and decisions, and most of that knowledge is scattered across documents and people. When it cannot be found, it is repeated, and when a person leaves, it is lost.
The cost is concrete. Teams re-run experiments that were already done, revisit questions that were already answered, and lose the reasoning behind past decisions when the people who made them move on. This is the tribal knowledge problem, and it compounds negatively as an organization grows. This article explains how AI-powered knowledge management changes that, and how internal knowledge becomes most valuable when connected to the external research record.
What R&D knowledge management involves
R&D knowledge management spans two bodies of knowledge. The first is internal: research reports, experimental results, technical decisions, and the reasoning behind them. The second is external: the patents, scientific literature, and competitive activity that place internal work in context. The goal is to make both retrievable in a way that reflects how researchers actually think about a problem, rather than by filename or folder.
The defining requirement is retrieval by meaning. A researcher rarely knows the exact document title or keyword; they know the problem. Knowledge management is only useful if a question about a compound, a method, or a decision returns the relevant internal work regardless of how it was originally labeled.
Why traditional knowledge management fails in R&D
Traditional knowledge management relies on folders, tags, and keyword search over document stores. It fails in R&D for the same reasons keyword search fails elsewhere: the same concept is described in different words across teams and years, so a query built on expected terms misses relevant work. Documents are siloed by team and system, and the connection between a past experiment and a current question is invisible.
It also fails at the human boundary. When knowledge lives in individuals rather than a retrievable system, staff turnover erases it. A traditional document repository preserves files but not the ability to find the right one at the right moment, which is the part that actually matters.
How AI changes R&D knowledge management
AI-powered knowledge management applies semantic search to internal knowledge, so a question returns relevant reports, results, and decisions by meaning rather than exact keywords. An R&D ontology organizes that knowledge by technical concept and connects related work, so a current problem surfaces the past work that bears on it even when the vocabulary differs.
The larger shift is connecting internal knowledge to the external record. When internal research is organized in the same conceptual structure as the external patent and scientific literature, a single question can reach both: what the team already knows, and what the wider field has published or patented. That connection is what turns a static archive into an intelligence layer.
Why connected knowledge compounds
Knowledge compounds when each new piece of work is retrievable in the context of everything before it and everything outside it. An experiment recorded today becomes findable the next time a related question arises; a past decision retains its reasoning; a current program is checked against both internal history and the external landscape before resources are committed. Instead of decaying as people leave and volume grows, the organization's knowledge becomes more valuable over time.
This is the difference between storing knowledge and compounding it. Storage preserves documents; compounding makes the whole body of work usable on every new question.
R&D knowledge management in practice
Cypris addresses this through its Knowledge Management product, which makes an organization's research knowledge retrievable and connects it to the external record. Internal work is organized through the same proprietary R&D ontology that structures a corpus of more than 500 million patents and scientific papers, so a single semantic query reaches both internal knowledge and the external patent and scientific literature.
Cypris Q, the platform's agentic layer, lets teams interrogate that combined knowledge in natural language and returns cited output, so a question about a compound or a program draws on internal history and external context at once. Cypris operates under enterprise API partnerships with OpenAI, Anthropic, and Google, with enterprise-grade security, and serves hundreds of enterprise customers across pharmaceuticals, chemicals, advanced materials, and other regulated industries.
FAQ
What is R&D knowledge management?
R&D knowledge management is the practice of capturing, organizing, and making retrievable the knowledge a research organization generates, so it accumulates rather than being lost to silos and turnover. It covers internal reports, experiments, and decisions, and connects them to the external patent and scientific record.
Why does R&D lose institutional knowledge?
R&D loses institutional knowledge because much of it lives in individuals and scattered documents rather than a retrievable system, so it disappears when people leave or when work cannot be found. This tribal knowledge problem leads teams to repeat experiments and lose the reasoning behind past decisions.
Why does traditional knowledge management fail in R&D?
Traditional knowledge management fails in R&D because folder-and-keyword systems miss work described in different terms across teams and years, and they silo documents by system. They preserve files but not the ability to find the right one at the right moment, which is the part that matters.
How does AI improve R&D knowledge management?
AI improves R&D knowledge management by applying semantic search, so a question returns relevant internal work by meaning rather than exact keywords. An R&D ontology organizes knowledge by technical concept and connects related work, and links internal knowledge to the external patent and scientific record.
What is tribal knowledge and why does it matter?
Tribal knowledge is the undocumented understanding held by individuals in an organization, such as why a decision was made or how a method actually works. It matters because it is lost when people leave, and capturing it in a retrievable system is a central goal of R&D knowledge management.
How does knowledge management connect internal work to external research?
Knowledge management connects internal work to external research by organizing both in the same conceptual structure, so a single question reaches internal reports and the external patent and scientific literature together. This places a team's own work in the context of what the wider field has published or patented.
What does it mean for knowledge to compound?
Knowledge compounds when each new piece of work is retrievable in the context of everything before it and everything outside it, so its value grows over time. Instead of decaying as staff turn over and volume rises, the organization's body of work becomes more usable on every new question.
Is R&D knowledge management just a document repository?
R&D knowledge management is more than a document repository, because storage alone preserves files without making the right one findable at the right moment. The value is in retrieval by meaning and in connecting internal knowledge to the external record, not in archiving.
Which teams benefit most from R&D knowledge management?
Research-intensive organizations benefit most from R&D knowledge management, particularly in pharmaceuticals, chemicals, advanced materials, and energy, where programs are long, knowledge is technical, and turnover erases hard-won understanding. These teams gain the most from preserving and connecting institutional knowledge.
What is the best platform for R&D knowledge management?
The best platform for R&D knowledge management makes internal knowledge retrievable by meaning and connects it to the external research record. Cypris does this through its Knowledge Management product, organizing internal work through the same R&D ontology that structures a corpus of more than 500 million patents and scientific papers.
mRNA therapeutics have moved from pandemic response to a broad modality, and their patent landscape is distinctive because the mRNA molecule is patented separately from the lipid nanoparticle that delivers it. This article addresses the construct itself. A therapeutic mRNA is engineered in several parts, and each is a distinct region of patenting: the modified nucleosides, such as pseudouridine variants, that reduce the innate immune response, an insight foundational to making mRNA usable in humans;¹ the five-prime cap that enables translation;² the untranslated regions that tune expression;³ the poly-A tail that stabilizes the molecule and, with the cap, defines the ends that are engineered for therapeutic performance;⁴ the sequence and codon optimization that improves output; and, increasingly, the self-amplifying and trans-amplifying designs that let a smaller dose replicate inside the cell. Rational-design analyses describe the construct as exactly this layered assembly, from cap through untranslated regions and open reading frame to poly-A tail, and the choice of nucleoside modification continues to shape immunogenicity.⁵,⁶ Because these elements can be claimed independently and are often held by different owners, and because delivery adds its own separate estate, freedom-to-operate for an mRNA product is a multi-layer, multi-owner analysis rather than a single clearance.
The field's IP has been defined by landmark disputes, which have raised the stakes across every layer. Foundational modified-nucleoside discoveries originated in academic work and are licensed through a chain of sublicenses, and the leading commercial developers have litigated over who owns and who may use the core construct technologies. Several of these cases have moved through courts in the United States and through the European Patent Office: one developer's settlement arrangements to resolve pending mRNA vaccine litigation with two others were entered on August 7, 2025,⁷ while a series of European construct patents were revoked or narrowed in opposition proceedings, with a further opposed patent maintained in amended form.⁸ The practical result is that a developer can hold a strong position on its own sequence and still face freedom-to-operate exposure on the nucleoside chemistry, the untranslated regions, or the tail, plus the separate delivery layer. This shows in the record: across the Cypris corpus of more than 500 million patents and scientific papers, the mRNA vaccine and therapeutics set holds on the order of 16,694 families and grew from about 582 in 2020 to roughly 2,062 in 2024, with the most active assignees including ModernaTx, Translate Bio, CureVac, the University of Pennsylvania, MIT, and BioNTech, and the United States far ahead of China and Germany on geography; 2025 and 2026 counts are partial because of the publication lag.
The strategic picture turns on where defensible, hard-to-design-around IP sits. The foundational modified-nucleoside and core-structure estates are comparatively crowded and heavily licensed, so the open, high-value ground is increasingly in self-amplifying and trans-amplifying mRNA, in novel nucleoside modifications and sequence-engineering methods, in untranslated-region and structural designs that improve durability and expression, in enzymatic capping and manufacturing methods, and in mRNA applications beyond vaccines.⁷ Self-amplifying mRNA has now reached the market: the first self-amplifying mRNA vaccine, which encodes a replicase alongside the antigen so the molecule copies itself inside the cell, was approved in Japan in 2023 and by the European Commission in February 2025, though not, as of this writing, in the United States.⁹ Circular RNA and other next-generation constructs are a further frontier. Reading the landscape by construct layer and by owner, and tracking both the patents and the underlying RNA-biology research, is what separates a workable position from a blocked one.
What creates FTO risk in mRNA constructs
Modified-nucleoside claims. These cover the chemistries, such as pseudouridine variants, that reduce immune activation, a foundational and heavily licensed layer.¹,⁶
Cap and untranslated-region claims. These cover the five-prime cap and the untranslated regions that enable and tune translation, a distinct expression-control layer.²,³
Poly-A tail and stability claims. These cover the tail and other end-engineering technologies, a separately owned and actively litigated layer.⁴
Sequence and codon-optimization claims. These cover methods to improve protein output from a given sequence, which can carry their own IP.
Self-amplifying and next-generation-construct claims. These cover self-amplifying, trans-amplifying, and circular-RNA designs, an emerging and comparatively open layer.⁹
How AI-powered landscape and FTO analysis helps
A modular, multi-owner, litigation-shaped landscape is beyond manual clearance. AI-powered analysis addresses this with semantic search that retrieves relevant nucleoside, cap, untranslated-region, tail, and self-amplifying claims regardless of terminology, attribution that resolves academic and commercial owners and the sublicense chains to canonical entities, claim-level analysis that separates the layers, and continuous monitoring that tracks new filings and disputes. Because RNA-biology advances appear in scientific literature before they are patented, reading both patents and literature gives earlier warning of where the field is heading.
Where Cypris fits
Cypris runs patent landscape and freedom-to-operate analysis for modular, contested fields such as mRNA therapeutics across a corpus of more than 500 million patents and scientific papers, organized through a proprietary R&D ontology. The ontology clusters the landscape by construct layer, modified nucleoside, cap, untranslated region, poly-A tail, and self-amplifying design, and normalizes academic and commercial owners and their sublicense chains to canonical entities, so a team sees how rights are distributed across the many parties rather than a flat list, and can separate the construct estate from the delivery estate. Semantic search across patents and scientific literature surfaces relevant claims regardless of terminology and connects filings to the underlying research, which is where new modifications and constructs emerge first. Cypris Q, the platform's agentic layer, lets teams run landscape and FTO analysis conversationally and chain the attribution, clustering, and claim-level analysis across layers, and Agentic Monitoring tracks the landscape over time and flags new filings and developments as they publish. Cypris provides enterprise API partnerships with OpenAI, Anthropic, and Google, and is built with enterprise-grade security. Cypris serves hundreds of enterprise customers across pharmaceuticals, chemicals, advanced materials, energy, and other regulated industries.
FAQ
Why is mRNA construct IP separate from delivery IP? mRNA construct IP is separate from delivery IP because the mRNA molecule and the lipid nanoparticle that carries it are distinct inventions with distinct owners. The construct covers the nucleosides, cap, untranslated regions, tail, and sequence; the delivery covers the lipid formulation. Freedom-to-operate must clear both estates separately.
Why were modified nucleosides so important? Modified nucleosides were important because replacing a natural nucleoside with a modified form, such as a pseudouridine variant, sharply reduced the innate immune response that had previously made mRNA unsuitable as a drug. This breakthrough helped enable therapeutic mRNA. The foundational nucleoside estates are therefore central to the landscape.
What claim types create FTO risk in mRNA constructs? Five claim types create FTO risk: modified-nucleoside claims, cap and untranslated-region claims, poly-A tail and stability claims, sequence and codon-optimization claims, and self-amplifying and next-generation-construct claims. Each covers a distinct layer and can be held by a different owner. The nucleoside and stability layers have been especially contested.
Why has mRNA IP been so heavily litigated? mRNA IP has been heavily litigated because the modality became commercially enormous very quickly, foundational construct technologies are held by a small number of parties, and their scope overlaps. Disputes have run through courts and the European Patent Office, with some resolved by settlement, including arrangements entered in August 2025, and others contested in opposition proceedings. The outcomes shape licensing across the field.
Has a self-amplifying mRNA product been approved? Yes. The first self-amplifying mRNA vaccine, which encodes a replicase so the mRNA copies itself inside cells, was approved in Japan in 2023 and by the European Commission in February 2025. It had not been approved in the United States as of this writing. Self-amplifying designs aim to achieve a given effect at a lower dose.
Where is the white space in mRNA therapeutics? The white space includes self-amplifying and trans-amplifying mRNA, novel nucleoside modifications and sequence engineering, untranslated-region and structural designs, enzymatic capping and manufacturing, circular RNA, and applications beyond vaccines. The foundational construct layers are crowded and licensed. The durable, defensible value is in next-generation constructs and new applications.
What software helps analyze the mRNA therapeutics patent landscape? Software for the mRNA landscape should resolve academic and commercial owners and sublicense chains to canonical entities, separate the construct and delivery estates, cluster the nucleoside, cap, untranslated-region, tail, and self-amplifying layers, search patents and scientific literature semantically, and monitor disputes and new filings continuously. Cypris does this across more than 500 million patents and scientific papers using a proprietary R&D ontology, semantic search, Cypris Q, and Agentic Monitoring.
Which teams need mRNA patent landscape and FTO analysis? mRNA patent landscape and FTO analysis is needed by R&D, IP, and business-development teams at mRNA and vaccine companies, as well as investors assessing mRNA assets. The modular, litigation-shaped landscape makes structured analysis essential. Cypris serves hundreds of enterprise customers across pharmaceuticals and other research-intensive industries.
Endnotes
- Karikó, K., Buckstein, M., Ni, H., & Weissman, D. (2005). Suppression of RNA recognition by Toll-like receptors: the impact of nucleoside modification and the evolutionary origin of RNA. Immunity, 23(2). https://doi.org/10.1016/j.immuni.2005.06.008
- Kore, A. R., Senthilvelan, A., & Shanmugasundaram, M. (2022). Recent advances in modified cap analogs for mRNA-based vaccines. The Chemical Record, 22(9). https://doi.org/10.1002/tcr.202200005
- Zhang, H., et al. (2024). Optimization of the 5′ untranslated region of mRNA vaccines. Scientific Reports, 14. https://doi.org/10.1038/s41598-024-70792-x
- Jemielity, J., et al. (2023). Chemical modifications of mRNA ends for therapeutic applications. Accounts of Chemical Research, 56(20). https://doi.org/10.1021/acs.accounts.3c00442
- To, K. K. W., & Cho, W. C. S. (2021). An overview of rational design of mRNA-based therapeutics and vaccines. Expert Opinion on Drug Discovery, 16(11). https://doi.org/10.1080/17460441.2021.1935859
- Liu, Y. (2026). The impact of nucleotide modifications on the immune responses of mRNA vaccines. https://doi.org/10.54097/bmddk068
- CureVac N.V. (2025). CureVac announces resolution of patent litigation with Pfizer/BioNTech (Form 6-K, Exhibit 99.1). U.S. Securities and Exchange Commission. https://www.sec.gov/Archives/edgar/data/1809122/000110465925075352/tm2522930d1_ex99-1.htm
- CureVac N.V. (2025). CureVac receives positive validity decision from the European Patent Office in litigation against BioNTech SE (Form 6-K, Exhibit 99.1). U.S. Securities and Exchange Commission. https://www.sec.gov/Archives/edgar/data/1809122/000110465925028798/tm2510706d1_ex99-1.htm
- European Medicines Agency (2024). Kostaive (zapomeran): EPAR public assessment report. https://www.ema.europa.eu/en/documents/assessment-report/kostaive-epar-public-assessment-report_en.pdf
Antibody-drug conjugates are among the most active areas of oncology drug development, and their patent landscape is distinctive because an ADC is a modular product whose components are patented separately. An ADC joins a monoclonal antibody to a cytotoxic payload through a chemical linker, using a defined conjugation chemistry and a specified drug-to-antibody ratio. Each of these elements, the antibody, the linker, the payload, the conjugation site and chemistry, and the ratio, can be claimed independently, so freedom-to-operate risk is layered across several distinct patent families held by different owners. Freedom-to-operate determines whether making, using, or selling a product would infringe another party's active patent claims, and peer-reviewed analysis of ADC intellectual property has long stressed that the assessment must cover every layer, not the molecule as a whole.¹
The landscape has grown intensely. A peer-reviewed update to the ADC patent literature notes that, a decade after the first ADC patent-landscape review, the basic principles still apply but the field has expanded and matured substantially, with next-generation payloads, linkers, and site-specific conjugation driving new filings.² That expansion is visible in the patent record: across the Cypris corpus of more than 500 million patents and scientific papers, ADC-specific patent families more than doubled from about 2,645 in 2018 to about 5,949 in 2024, with 2025 counts partial because of the roughly eighteen-month publication lag. The growth has been propelled by potent topoisomerase-1 payloads such as the deruxtecan and govitecan classes, new linker and site-specific conjugation technologies, and the expansion of ADCs from hematologic cancers into solid tumors. Peer-reviewed patent reviews map the issued patents onto specific linker and payload technologies,³ and document filing activity concentrated among a small set of leading developers, with more than a dozen approved ADCs and a large clinical pipeline behind the trend.⁴ Within the Cypris corpus, conjugation and site-specific chemistry and the linker layer are the most heavily worked parts of the ADC set, consistent with where litigation and FTO risk concentrate.
Litigation has made the stakes concrete, and it has centered on the linker layer. In the multi-year dispute between Seagen and Daiichi Sankyo over the linker technology used in a blockbuster HER2-targeted ADC, a jury had found for Seagen and awarded damages, but on December 2, 2025 the US Court of Appeals for the Federal Circuit reversed, holding Seagen's key linker patent invalid for lack of written description and non-enablement and vacating the damages award.⁵ The court reasoned that the priority disclosure did not convey possession of the specific claimed subgenus of linkers and that a broad functional claim was not enabled.⁵ For developers, the practical lesson is twofold: the linker and conjugation layer is heavily contested and a frequent source of FTO risk, and the proprietary payload estates built around leading platforms, such as the DXd payload, create freedom-to-operate exposure for follow-on and biosimilar ADCs in markets where those estates are in force. That exposure is concentrated: across the Cypris corpus, the most active assignees in the ADC-specific set include Genentech, Seagen, Daiichi Sankyo, Regeneron, Immunomedics, and ImmunoGen, several of which anchor the payload and linker estates most likely to surface in an FTO search. Because applications publish about eighteen months after filing, the newest linker, payload, and conjugation filings are under-represented, so the current frontier is more active than granted-patent counts suggest.
What creates FTO risk in ADCs
Antibody claims. These cover the targeting antibody and its engineering, a distinct layer that can implicate separate antibody IP.
Linker claims. These cover cleavable and non-cleavable linkers and their chemistry, the layer most heavily litigated, as the Seagen v. Daiichi Sankyo dispute demonstrates.⁵
Payload claims. These cover the cytotoxic agent, including proprietary payload estates built around specific classes, which create FTO exposure for follow-on products.
Conjugation and site-specific claims. These cover how payload and antibody are joined and where, an area of intense recent innovation and patenting.³
Drug-to-antibody ratio and formulation claims. These cover the ratio and the finished formulation, adding further independently claimable layers.
How AI-powered landscape and FTO analysis helps
A modular, multi-owner, actively litigated landscape is beyond manual clearance. AI-powered analysis addresses this with semantic search that retrieves relevant antibody, linker, payload, and conjugation claims regardless of terminology, attribution that resolves the many owners to canonical entities, claim-level analysis that separates the layers, and continuous monitoring that tracks new filings and litigation developments. Because ADC advances appear in scientific literature before they are patented, reading both patents and literature gives earlier warning of where the landscape is extending.
Where Cypris fits
Cypris runs patent landscape and freedom-to-operate analysis for modular, contested fields such as antibody-drug conjugates across a corpus of more than 500 million patents and scientific papers, organized through a proprietary R&D ontology. The ontology clusters the landscape by layer, antibody, linker, payload, and conjugation, and normalizes owners to canonical entities, so a team sees how rights are distributed across the many parties rather than a flat list. Semantic search across patents and scientific literature surfaces relevant claims regardless of terminology and connects filings to the underlying research, which is where next-generation linkers and payloads emerge first. Cypris Q, the platform's agentic layer, lets teams run landscape and FTO analysis conversationally and chain the attribution, clustering, and claim-level analysis across layers, and Agentic Monitoring tracks the landscape over time and flags new filings and developments as they publish. Cypris provides enterprise API partnerships with OpenAI, Anthropic, and Google, and is built with enterprise-grade security. Cypris serves hundreds of enterprise customers across pharmaceuticals, chemicals, advanced materials, energy, and other regulated industries.
FAQ
Why is freedom-to-operate hard for antibody-drug conjugates? Freedom-to-operate is hard for antibody-drug conjugates because an ADC is a modular product whose antibody, linker, payload, conjugation chemistry, and drug-to-antibody ratio are each independently patentable and often held by different owners. Clearing one layer does not clear the others. FTO must therefore be assessed layer by layer across multiple patent families.
What are the main claim types in the ADC landscape? The main claim types are antibody claims, linker claims, payload claims, conjugation and site-specific claims, and drug-to-antibody-ratio and formulation claims. Each covers a distinct layer of the ADC and can independently create infringement risk. The linker and conjugation layers are especially heavily patented and litigated.
What was the Seagen v. Daiichi Sankyo dispute about? The Seagen v. Daiichi Sankyo dispute concerned linker technology used in a blockbuster HER2-targeted ADC. A jury had found for Seagen and awarded damages, but on December 2, 2025 the US Court of Appeals for the Federal Circuit reversed, holding Seagen's key linker patent invalid for lack of written description and enablement and vacating the award. It illustrates how the linker layer drives ADC freedom-to-operate risk and how even a trial win can be undone on validity grounds.
How fast is ADC patenting growing? ADC patenting has grown rapidly, with ADC-specific patent families more than doubling between 2018 and 2024 in the Cypris corpus. Growth has been driven by potent topoisomerase-1 payloads, new linker and site-specific conjugation technologies, and expansion from hematologic cancers into solid tumors. Because applications publish about eighteen months after filing, recent activity is under-represented.
What is a payload estate and why does it matter for FTO? A payload estate is the set of patents an organization holds around a specific cytotoxic payload class and its use in ADCs. It matters for FTO because a strong payload estate, such as the one around the DXd payload, can create infringement exposure for follow-on and biosimilar ADCs in markets where it is in force. Developers must assess payload IP as a distinct layer.
Why does ADC analysis need scientific literature? ADC analysis needs scientific literature because linker, payload, and conjugation advances appear in research before they are patented, so the literature gives the earliest signal of where the landscape is extending. Analyzing patents alone gives a lagging view. Cypris analyzes both across more than 500 million patents and scientific papers.
Which teams need ADC patent landscape and FTO analysis? ADC patent landscape and FTO analysis is needed by R&D, IP, and business-development teams at pharmaceutical and biotech companies developing ADCs, payloads, linkers, and conjugation platforms, as well as investors assessing ADC assets. The modular, litigated landscape makes structured analysis essential. Cypris serves hundreds of enterprise customers across pharmaceuticals and other research-intensive industries.
How current does an ADC landscape need to be? An ADC landscape needs to be continuously current, because litigation is active, next-generation linker and payload filings publish constantly, and publication lag hides the most recent activity. A one-time landscape ages quickly. Cypris uses Agentic Monitoring to track the landscape and flag new filings and developments as they publish.
Battery circularity, the recycling, reuse, and repurposing of batteries, has become the fastest-growing area of battery patenting, and its landscape is a map of the coming competition over critical minerals. According to a joint study by the European Patent Office and the International Energy Agency, international patent families related to battery circularity grew at an average of 42 percent per year from 2017 to 2023, compared with 16 percent for rechargeable battery manufacturing overall and 2 percent across all technical fields.¹ Over the decade the number of these families rose roughly sevenfold.¹ The driver is structural: more than one in four cars sold globally in 2025 was electric, and around 1.2 million electric-vehicle batteries could reach end of life in 2030, rising to 14 million by 2040, so managing and reclaiming that volume is both an environmental necessity and a supply-chain strategy.¹
The landscape is geographically concentrated and shifting quickly. Asian applicants accounted for 63 percent of battery-circularity patent families in 2023, and China's share rose from 5 percent in 2013 to 29 percent in 2023, with Brunp, the recycling subsidiary of a major battery maker, overtaking established Japanese and Korean firms to become the most active filer.¹ European companies and research institutes account for roughly 20 percent of families, with particular strength in the collection and pre-processing of used batteries and in chemical transformation to recover raw materials, reflecting Europe's current role more as a battery user than a producer.¹ An independent count across the Cypris corpus of more than 500 million patents and scientific papers reproduces the same picture: China holds roughly two-thirds of the recycling-specific family set, well ahead of the United States, Germany, South Korea, and Japan, and Brunp is the single most active assignee, ahead of chemical and battery-materials firms such as BASF and Sumitomo Metal Mining. The strategic significance is large: energy storage now represents about 40 percent of all energy-related patenting and is heading toward half, and recycled materials could meet more than a fifth of demand for lithium, nickel, and cobalt by 2040.¹
The technology landscape divides into distinct stages, each a region of patenting. A peer-reviewed patent-network analysis of lithium-ion battery recycling covering 1990 to 2024 finds activity rising steeply since around 2020, with China leading and international collaboration remaining limited,² and bibliometric analysis of the field documents the same long-run acceleration in recycling research and patenting.³ Across the Cypris corpus, hydrometallurgy is the most patented chemical-recovery route, well ahead of pyrometallurgy, while direct recycling and cathode regeneration remain comparatively nascent; a large, separate cluster covers collection, pre-processing, and separation, the earlier stage where Europe is comparatively strong. Metal recovery and cathode regeneration are where much of the chemical innovation and value concentrate, with key work focused on improving leaching efficiency, developing purification methods, and relithiation strategies that restore spent cathode materials. Because applications publish about eighteen months after filing, the most recent activity is under-represented, so the current frontier is even more active than the figures show.
Where the battery-circularity white space is
Direct cathode regeneration. Restoring spent cathode material directly, rather than breaking it down to metals, is a higher-value route that remains comparatively nascent in the patent record, leaving room for defensible positions.²
Efficient metal recovery. Improving leaching efficiency and purification for lithium, nickel, and cobalt is where much chemical innovation concentrates and where recovery economics are decided.²
Collection and pre-processing. Sorting, dismantling, and safe handling, including remote-handling technologies, are an earlier stage where activity is comparatively less crowded and where European applicants are relatively strong.¹
Reuse and repurposing. Second-life applications for batteries, distinct from material recovery, are a separate and growing layer.
Design for recyclability. Battery designs that ease disassembly and recovery link circularity back to manufacturing and are an emerging cross-over area.
How AI-powered landscape and white space analysis helps
Resolving a fast-growing landscape across stages, chemistries, and geographies requires more than keyword search. AI-powered analysis addresses this with semantic search that clusters activity by recovery route and processing stage across varied terminology, attribution that normalizes filers to canonical entities and tracks shifting leadership, and continuous monitoring that keeps pace with a field growing far faster than average. Because circularity advances appear in scientific literature before they are patented, reading both patents and literature gives the earliest signal of where the frontier is moving.
Where Cypris fits
Cypris runs patent landscape and white space analysis for fast-growing energy fields such as battery circularity across a corpus of more than 500 million patents and scientific papers, organized through a proprietary R&D ontology. The ontology clusters activity by recovery route, hydrometallurgy, direct regeneration, separation, and pyrometallurgy, and by processing stage, and normalizes filers to canonical entities, so a team can resolve which routes and stages are crowded and which remain open as white space, and can track shifting leadership as new entrants rise. Semantic search across patents and scientific literature connects filings to the underlying materials and process research, which is where circularity advances appear first. Cypris Q, the platform's agentic layer, lets teams run landscape and white space analysis conversationally and chain the clustering, attribution, and gap analysis, and Agentic Monitoring tracks a defined route over time and flags new patents and papers as they publish. Cypris provides enterprise API partnerships with OpenAI, Anthropic, and Google, and is built with enterprise-grade security. Cypris serves hundreds of enterprise customers across pharmaceuticals, chemicals, advanced materials, energy, and other regulated industries.
FAQ
How fast is battery-recycling patenting growing? Battery-recycling patenting is growing very fast. According to the EPO and IEA, international patent families in battery circularity grew at an average of 42 percent per year from 2017 to 2023, versus 16 percent for battery manufacturing and 2 percent across all technical fields, roughly a sevenfold increase over the decade. It is now growing faster than battery patenting in general.
Who leads in battery-circularity patents? Asian applicants held 63 percent of battery-circularity patent families in 2023. China's share rose from 5 percent in 2013 to 29 percent in 2023, and Brunp, a major battery maker's recycling subsidiary, overtook established Japanese and Korean firms as the most active filer. European companies and research institutes hold roughly 20 percent, with strength in collection and pre-processing. An independent Cypris-corpus count reproduces China's roughly two-thirds share and Brunp's lead.
What technologies does the battery-recycling landscape cover? The battery-recycling landscape covers collection, sorting, and dismantling; mechanical processing; and metal recovery and cathode regeneration. Analysis of the patent record finds hydrometallurgy the most patented chemical-recovery route, ahead of pyrometallurgy, with direct recycling still comparatively nascent. Innovation concentrates on leaching efficiency, purification, and relithiation.
Why is battery circularity strategically important? Battery circularity is strategically important because it is a secondary supply of critical minerals. Around 1.2 million electric-vehicle batteries could reach end of life in 2030 and 14 million by 2040, and recycled materials could meet more than a fifth of lithium, nickel, and cobalt demand by 2040. This links recycling to supply-chain security and energy security.
Where is the white space in battery recycling? The white space in battery recycling includes direct cathode regeneration, efficient metal recovery and purification, collection and pre-processing including remote handling, reuse and repurposing for second-life applications, and design for recyclability. Metal recovery and cathode regeneration are where chemical innovation concentrates. The higher-value opportunities are in routes that improve recovery economics.
Why does battery-recycling analysis need scientific literature? Battery-recycling analysis needs scientific literature because process and materials advances appear in research before they are patented, so the literature gives the earliest signal. Analyzing patents alone gives a lagging view. Cypris analyzes both across more than 500 million patents and scientific papers.
Which teams use battery-recycling patent landscape analysis? Battery-recycling patent landscape analysis is used by R&D, innovation, IP, and strategy teams at battery makers, recyclers, automotive and energy companies, materials developers, and their partners, as well as investors and policymakers. It informs where to invest, where to file, and where competitors are concentrated. Cypris serves hundreds of enterprise customers across energy, advanced materials, chemicals, and other regulated industries.
How do you keep a battery-recycling landscape current? Keeping a battery-recycling landscape current requires continuous monitoring, because the field is growing far faster than average, leadership is shifting quickly, and publication lag hides the most recent activity. A one-time landscape ages quickly. Cypris uses Agentic Monitoring to track a defined route and flag new patents and papers as they publish.
Endnotes
- International Energy Agency & European Patent Office (2026). Battery circularity: innovation trends for a future source of critical materials. IEA, Paris. https://www.iea.org/reports/battery-circularity
- von Delft, S., Schlehuber, S., & Hemmelder, A. (2025). Uncovering collaboration and knowledge areas in lithium-ion battery recycling. EES Batteries. https://doi.org/10.1039/d5eb00056d
- Li, Y., Guo, Y., Guan, J., Zhang, X., & Lou, X. (2022). Global trend for waste lithium-ion battery recycling from 1984 to 2021: a bibliometric analysis. Minerals, 12(12), 1514. https://doi.org/10.3390/min12121514
AAV gene therapy has crossed into commercial reality, and its patent landscape is distinctive because an AAV therapy is a modular product whose parts are patented separately. An adeno-associated virus vector delivers a therapeutic gene by packaging it inside an engineered protein shell, the capsid, whose surface engages target-cell receptors and whose fate through endocytosis, endosomal escape, and nuclear import determines where the therapy goes and how the immune system responds to it.¹ The intellectual property divides across distinct regions, each a distinct area of patenting: the capsid, spanning natural serotypes and, increasingly, engineered capsids produced by directed evolution, structure-guided design, and machine-learning-guided design;²,³,⁴,⁵ the strategies that address immunogenicity, because pre-existing neutralizing antibodies exclude many patients and the immune response generally prevents redosing;⁶,⁷ the transgene expression cassette, including the promoter and regulatory elements that control where and how strongly the gene is expressed; and the manufacturing process, where empty-capsid content, host-cell productivity, and the ratio of full to empty particles remain challenges at commercial scale.⁸ Because a therapy depends on several of these layers and they are often held by different owners, freedom-to-operate for an AAV product is a multi-layer, multi-owner analysis rather than a single clearance.
The competitive and legal environment has raised the stakes across every layer. Capsid engineering is the most active area of IP, with specialized platform companies developing next-generation capsids that target specific tissues such as the brain, muscle, and retina and that aim to evade pre-existing immunity, and licensing these platforms to larger developers.⁴,⁹ At the same time, foundational vector patents have been litigated, with the Federal Circuit deciding an appeal on core AAV vector claims in early 2026, so the boundaries of the foundational estate continue to be tested even as the field advances.¹⁰ The commercial ground is set by a broader wave of approved products: as of the 2024 review window, the US Food and Drug Administration had approved fourteen cellular and gene therapy products across the modality, of which the AAV-based Luxturna (voretigene neparvovec-rzyl) for RPE65-mediated inherited retinal dystrophy remains the AAV exemplar, first approved in the United States in 2017 and separately authorized in the European Union.⁴,¹¹,¹² Because applications publish about eighteen months after filing, the newest capsid and immune-evasion filings are under-represented, so the current frontier is more active than granted-patent counts suggest.
The strategic question is where defensible, hard-to-design-around IP sits. Across the Cypris corpus of more than 500 million patents and scientific papers, the AAV capsid-variant set holds on the order of 12,966 families and grew from about 828 in 2020 to roughly 1,818 in 2024, with the most active assignees led by the University of Pennsylvania, Voyager Therapeutics, the University of Massachusetts, Genzyme, and UC San Diego, and the United States far ahead of China, France, and the United Kingdom on geography; 2025 and 2026 counts are partial because of the publication lag. The most active and high-value ground is in engineered capsids that both target a tissue precisely and evade pre-existing immunity, because these directly address the field's central limitations, and receptor-guided and machine-learning-guided capsid design are advancing quickly here.²,³,⁴,⁵ Redosing and immune-modulation technologies are a distinct and comparatively open layer,⁶,⁷ as are manufacturing methods that raise the full-to-empty ratio and lower cost,⁸ and expression-cassette designs that improve durability and tissue specificity.⁴,⁹ Reading the landscape by layer and by owner, and tracking both the patents and the underlying virology and immunology research, is what separates a workable position from a blocked one.
What creates FTO risk in AAV gene therapy
Capsid claims. These cover natural serotypes and engineered capsids from directed evolution, structure-guided, and machine-learning design, the most active and contested layer.²,³,⁴,⁵
Immunogenicity and redosing claims. These cover strategies to evade pre-existing antibodies and enable redosing, a distinct and high-value layer given the field's central limitation.⁶,⁷
Transgene and expression-cassette claims. These cover promoters, regulatory elements, and the engineered transgene that control expression, a separately owned layer.
Manufacturing and purification claims. These cover producer systems, full-to-empty separation, host-cell productivity, and purification, where practical, hard-to-design-around barriers concentrate.⁸
Tissue-targeting claims. These cover receptor-guided and tissue-specific delivery, which can independently constrain a competing program.⁹
How AI-powered landscape and FTO analysis helps
A modular, multi-owner, litigation-shaped landscape is beyond manual clearance. AI-powered analysis addresses this with semantic search that retrieves relevant capsid, immunogenicity, cassette, and manufacturing claims regardless of terminology, attribution that resolves the many owners and license chains to canonical entities, claim-level analysis that separates the layers, and continuous monitoring that tracks new filings and disputes. Because AAV advances appear in virology and immunology literature before they are patented, reading both patents and literature gives earlier warning of where the field is heading.
Where Cypris fits
Cypris runs patent landscape and freedom-to-operate analysis for modular, contested fields such as AAV gene therapy across a corpus of more than 500 million patents and scientific papers, organized through a proprietary R&D ontology. The ontology clusters the landscape by layer, capsid, immunogenicity, expression cassette, and manufacturing, and normalizes developer, platform-specialist, and academic owners and their license chains to canonical entities, so a team sees how rights are distributed across the many parties rather than a flat list. Semantic search across patents and scientific literature surfaces relevant claims regardless of terminology and connects filings to the underlying research, which is where new capsids and immune-evasion strategies emerge first. Cypris Q, the platform's agentic layer, lets teams run landscape and FTO analysis conversationally and chain the attribution, clustering, and claim-level analysis across layers, and Agentic Monitoring tracks the landscape over time and flags new filings and developments as they publish. Cypris provides enterprise API partnerships with OpenAI, Anthropic, and Google, and is built with enterprise-grade security. Cypris serves hundreds of enterprise customers across pharmaceuticals, chemicals, advanced materials, energy, and other regulated industries.
FAQ
Why is freedom-to-operate hard for AAV gene therapy? Freedom-to-operate is hard for AAV gene therapy because a therapy is built from a capsid, an immune-evasion strategy, a transgene expression cassette, and a manufacturing process, each independently patentable and often held by different owners. The capsid layer alone is heavily engineered and contested. FTO must be assessed layer by layer across multiple estates.
Why is the capsid the key layer? The capsid is the key layer because it determines which tissues the therapy reaches and how the immune system responds, and it is where most engineering and patenting activity concentrates. Engineered capsids from directed evolution, structure-guided design, and machine learning aim to target tissues and evade immunity. That makes capsid IP the most active battleground.
What claim types create FTO risk in AAV therapy? Five claim types create FTO risk: capsid claims, immunogenicity and redosing claims, transgene and expression-cassette claims, manufacturing and purification claims, and tissue-targeting claims. Each covers a distinct layer and can be held by a different owner. The capsid and manufacturing layers are especially decisive.
Why do immunity and redosing matter so much? Immunity and redosing matter because pre-existing neutralizing antibodies exclude many patients from AAV therapy, and the immune response to a first dose generally prevents giving a second, so AAV is typically a single-dose modality. Technologies that evade pre-existing immunity or enable redosing address a central limitation. They are therefore a distinct, high-value layer.
Where is the white space in AAV gene therapy? The white space includes engineered capsids that target tissues and evade pre-existing immunity, redosing and immune-modulation technologies, manufacturing methods that raise the full-to-empty ratio, and expression-cassette designs that improve durability and specificity. Natural serotypes and liver-directed approaches are comparatively crowded. The durable, defensible value is in capsids, immune evasion, and manufacturing.
Why does AAV analysis need scientific literature? AAV analysis needs scientific literature because new capsids, immune-evasion strategies, and manufacturing advances appear in virology and immunology research before they are patented, so the literature gives the earliest signal. Analyzing patents alone gives a lagging view. Cypris analyzes both across more than 500 million patents and scientific papers.
What software helps analyze the AAV gene therapy patent landscape? Software for the AAV landscape should resolve developer, platform-specialist, and academic owners and license chains to canonical entities, cluster the capsid, immunogenicity, cassette, and manufacturing layers, search patents and scientific literature semantically, and monitor litigation and new filings continuously. Cypris does this across more than 500 million patents and scientific papers using a proprietary R&D ontology, semantic search, Cypris Q, and Agentic Monitoring.
Which teams need AAV patent landscape and FTO analysis? AAV patent landscape and FTO analysis is needed by R&D, IP, and business-development teams at gene therapy and pharmaceutical companies, capsid-platform specialists, academic technology-transfer offices, and investors assessing gene therapy assets. The modular, litigation-shaped landscape makes structured analysis essential. Cypris serves hundreds of enterprise customers across pharmaceuticals and other research-intensive industries.
Endnotes
- Hao, Y., & Xiang, J. (2023). Biophysical characterization of the AAV capsid through the viral transduction life cycle. Journal of Genetic Engineering and Biotechnology, 21. https://doi.org/10.1186/s43141-023-00518-5
- Fakhiri, J., Becker, S., & Grimm, D. (2022). Fantastic AAV gene therapy vectors and how to find them: random diversification, rational design and machine learning. Pathogens, 11(7), 756. https://doi.org/10.3390/pathogens11070756
- Qi, Y., et al. (2025). Artificial intelligence-based approaches for AAV vector engineering. Advanced Science, 12. https://doi.org/10.1002/advs.202411062
- Gao, F., et al. (2024). AAV engineering and load strategy for tropism modification, immune evasion and enhanced transgene expression. International Journal of Nanomedicine, 19. https://doi.org/10.2147/ijn.s459905
- Fu, W., et al. (2024). Machine-learning-guided directed evolution for AAV capsid engineering. Current Pharmaceutical Design. https://doi.org/10.2174/0113816128286593240226060318
- Barnes, C., Scheideler, O., & Schaffer, D. (2019). Engineering the AAV capsid to evade immune responses. Current Opinion in Biotechnology, 60. https://doi.org/10.1016/j.copbio.2019.01.002
- Meumann, N., Rodríguez-Márquez, E., & Büning, H. (2020). AAV capsid engineering in liver-directed gene therapy. Expert Opinion on Biological Therapy, 21(6). https://doi.org/10.1080/14712598.2021.1865303
- Yoon, S., Lee, D., et al. (2024). Decoding cellular mechanism of rAAV and engineering host-cell factories. Biotechnology Advances, 72. https://doi.org/10.1016/j.biotechadv.2024.108322
- Marković, I., et al. (2025). AAV gene therapy drug development and translation of engineered ocular and neurotropic capsids. Clinical and Translational Science, 18. https://doi.org/10.1111/cts.70428
- RegenxBio Inc. v. Sarepta Therapeutics, Inc., No. 24-1408 (Fed. Cir. Feb. 20, 2026).
- U.S. Food and Drug Administration. Approved cellular and gene therapy products. https://www.fda.gov/vaccines-blood-biologics/cellular-gene-therapy-products/approved-cellular-and-gene-therapy-products
- U.S. Food and Drug Administration (2017). Luxturna (voretigene neparvovec-rzyl). https://www.fda.gov/vaccines-blood-biologics/cellular-gene-therapy-products/luxturna
.avif)