P

PatSnap

A data platform that existed for over a decade before the models, with a substantial AI layer built on top. Founded in 2007, PatSnap's underlying asset is its data: more than 2 billion structured data points and over 210 million patent, science and technology records, with search, landscaping, dashboards and analytics that predate generative AI and work without it. The vendor itself describes Eureka as a capability layer rather than a separate platform, which is an accurate self assessment.

The AI layer is substantial rather than decorative, spanning agents for prior art, freedom to operate, drafting, translation and life sciences work, and the vendor describes a four stage data pipeline built to feed it. A mature platform hosts the model layer rather than depending on it. Remove the models and PatSnap remains a working patent analytics business.

PatentBench is published as a benchmark built for real patent work, measuring specific tasks against structured data, legal context and source linked evidence, with a full methodology the vendor links to. The published results are a 77 percent hit rate on design freedom to operate and a claim of being 83 times more accurate than ChatGPT on that measure. Naming the comparator and the task means an outsider can find the methodology and disagree with it without contacting the vendor.

Grounding is also described in the architecture: retrieval augmented generation over a governed corpus, a four stage pipeline the vendor says reduces hallucinations, and source linked evidence as a property of the benchmark. Three limits apply. The benchmark is the vendor's own rather than independent. The 83 times figure compares against a general purpose chatbot on a task it was never built for, which flatters the result. The methodology page itself was not opened and may say more.

Agent language is used throughout and no oversight model is published. The platform is described as agentic across IP and life sciences, with named agents including an FTO Agent, a lead compound agent, an Antibody Target Predictor and a Markush Claim Drafting agent. A published customer account describes completing infringement analysis for new products in hours rather than weeks. Data governance is described at each stage of the pipeline, which governs inputs rather than agent behavior.

Nothing published states what an agent may complete without human review, whether any output can be relied on without verification, what confidence signal accompanies an agent result, or where a human checkpoint sits in a workflow that runs from query to report. The gap matters because the outputs are decision grade: a freedom to operate assessment that clears a product for launch, and a novelty judgment that decides whether a molecule is pursued.

Named customers, a named executive and independent coverage are all dated and checkable. Named customers include NASA, Tesla, Vodafone, MIT and General Electric, reported in independent trade coverage and not only on vendor pages. NCH Corporation is also named, with Michael Schuster, VP of Product Development and Innovation, quoted by name and title. Corporate history is public: founded in 2007, a unicorn valuation in 2021 after a $300m Series E led by SoftBank Vision Fund 2 and Tencent, and Jeffrey Tiong as chief executive.

Independent evidence goes beyond press releases: a Gartner Peer Insights listing, a SoftwareOne marketplace listing, and a hands on review by a named trade publication that took a trial and published its assessment. An outsider can verify the company, the customers, the funding and the product without contacting the vendor. The outcome claims, productivity up to 60 percent higher and workflows 90 percent faster, carry no method, and one customer quote about hundreds of thousands of dollars saved names no organization.

General enterprise confidentiality is evidenced and the legal professional dimension is absent. Published: strict data isolation, ISO certified security policies, encryption in transit and at rest, consent management, GDPR and CCPA compliance, and positioning as a privacy first partner for enterprises. Those are real controls over the confidential research and portfolio material customers upload. Nothing mentions attorney client privilege or work product, or acknowledges that a law firm user's material carries professional obligations distinct from a corporate research team's. The buyer profile explains the gap without excusing it.

PatSnap sells across research and development, strategy, life sciences and legal, and treats confidentiality as an enterprise data question throughout. That is coherent for its market, and it leaves the practitioner question unanswered.

Not located. The platform generates freedom to operate assessments, novelty and design patent risk analyses, and drafts patent claims including Markush structures, all outputs that a practitioner signs off or a business relies on. The published customer accounts describe research and development staff who are not lawyers using them for risk decisions. Nothing published addresses whether output is legal advice, what role qualified counsel plays where an engineer or chemist runs a freedom to operate agent directly, or any professional conduct framework. The question is pressing for this vendor because its named buyers include scientists and engineers, not only attorneys.

Data governance is described in operational detail and model governance is not. The vendor publishes that a four stage pipeline processes over 2 billion data points across 20 domains with rigorous governance at every stage, says this is what reduces hallucinations and improves reliability, and names its retrieval architecture. PatentBench is a published evaluation instrument, and its existence shows that measurement happens.

Governance of the models themselves is absent. There is no AI policy, no model card, no bias or fairness testing, no accuracy monitoring beyond the benchmark, no drift statement, no named governance body, no ISO 42001 and no EU AI Act positioning. The vendor documents the governance of its data thoroughly and the governance of its system not at all.

Isolation and compliance are stated, and the training question is unanswered. Published: strict data isolation, ISO certified security policies, encryption in transit and at rest with named protocol and cipher, GDPR and CCPA compliance with stated consent management, and a self description as privacy first. Nothing is published on whether documents customers upload, search histories, proprietary formulations or research materials are used to train or improve the domain models, and no retention period or deletion right is stated.

The gap is material because customers upload their own novel compounds, formulations and unpublished technical material, which is competitively sensitive in a way patent search over public data is not.

No published position is published on liability for AI output, warranty, service levels or remedy. The exposure here falls on business decisions rather than filings. A published customer account describes the FTO Agent completing infringement analysis in hours, and another describes lead compound screening as the first gate in a research process, so a false clearance sends a product to market and a false negative on novelty kills a viable molecule.

The vendor publishes a 77 percent hit rate on design freedom to operate, which by its own arithmetic means roughly one in four is missed, and no published position says who carries that.

The Patsnap Open Platform is a dedicated developer product rather than a connector list, with four integration surfaces: REST APIs, MCP servers, UI widgets and agent skills, backed by a published Developer Center and documentation. Coverage is counted: more than 20 patent APIs, more than 25 corporate APIs and more than 16 infringement APIs. The MCP servers expose the data and its agents as tools for outside AI systems, and the vendor names the frameworks it targets, including LangChain, AutoGen and function calling workflows.

Agent skills are also published as a surface. A buyer can get an API key from the site, so a developer can assess, obtain and build against the platform without contacting sales. No named IP management or docketing system integration is published, which is the connection a prosecution practice would ask for first.

Nothing located. No hosting provider is named, no region or residency commitment is published, and no single tenant, dedicated or on premises option is described. The absence matters here. The company is headquartered in Singapore, names SoftBank and Tencent among its investors and serves customers including NASA, MIT and General Electric, so where research material and unpublished technical data are processed is a question those customers' compliance functions would ask directly, and public material does not answer it. Stated GDPR and CCPA compliance shows that those regimes are addressed and says nothing about where data sits.

A published trust center names standards and annual audits, and the SOC report is Type 1. The trust center lists certification by the AICPA, ISO 27001:2022 with a statement that systems undergo annual external audits, TLS 1.3 and AES-256 encryption in transit and at rest, GDPR and CCPA compliance, and consent management. Naming the ISO revision year and committing to annual external audit addresses currency. The SOC attestation is announced and described as SOC 2 Type 1.

Type 1 assesses the design of controls at a single point in time, while Type 2 tests whether those controls operated effectively over a period, and the difference is not cosmetic. No auditing firm is named, and no certificate date is given for the ISO certification.

Nothing located. No foundation model provider, model family or version is named, and no subprocessor list was found. The vendor describes its architecture in technical terms, naming retrieval augmented generation and what it calls retrieval augmented thinking, and describes domain specific and domain trained models, all of which characterize method rather than provenance. Independent product review notes an evolving AI backbone, which implies underlying models change over time and makes identification more rather than less relevant for a buyer assessing consistency.

For a platform whose Open Platform exposes agents to external systems through MCP servers, the identity of the models behind those agents is a question a developer integrating them would ask.

The pricing structure is disclosed and no figure is. An independent analyst platform, consistent with vendor material, describes a subscription with multiple tiers that vary by features, user numbers and access level, with details given on request, and a named Enterprise plan that includes agents such as the Antibody Target Predictor and Markush Claim Drafting. Knowing that the meter runs on seats and feature tier, and which capabilities sit in the highest plan, tells a buyer how the bill will grow and which plan they would need.

Two self serve entry points lower the barrier: a free trial of Eureka and an API key from the Open Platform. No price, range or entry figure is published at any level, and every commercial route ends in a quote request. That structure comes from the analyst listing because no vendor pricing page is published.

Coverage is stated with counts and on the dimensions a researcher checks. Scale: more than 2 billion structured data points and over 210 million patent, science and technology records, spanning patents, scientific literature, litigation records, clinical data and technical knowledge across 20 domains. Coverage extends beyond patents, which matters because prior art is not confined to patents. Industries are listed: agriculture and chemicals, consumer goods, food and drink, life sciences, automotive, oil and gas, professional services, aviation and aerospace, and education.

Depth in life sciences is shown through named capabilities for sequences, Markush structures, structure activity relationship extraction, clinical trial comparison and druggability. Patent translation trained on global patent data addresses work across jurisdictions. A buyer can test these counts, domains and named document types against their own work. No update lag or refresh frequency is published, and no list of patent offices is given.