PatentWatch
PatentWatch is an AI platform for finding out who is infringing a patent portfolio and building the evidence to act on it. Its four published modules are product mapping, patent search, prior art analysis and portfolio analytics. The core workflow maps granted claims against competing products at scale: the platform interprets a claim, breaks it into its elements, locates supporting evidence in product documentation and specifications, and assembles a claim chart in which each element carries the passage relied on, an analysis and a numeric score, with an overall match score for the claim. Prior art and invalidity searching runs alongside it, aimed at testing whether a patent would survive a challenge. The company describes the intended uses as identifying licensing and litigation targets, supporting cease-and-desist correspondence, finding counter-claims when a company is itself sued, valuing intellectual property in wind-downs, and accelerating diligence in mergers and acquisitions. PatentWatch positions the product against manual claim charting by attorneys and against in-house teams able to monitor only part of a portfolio, and frames the opportunity as portfolios that cost a great deal to obtain and maintain while generating no revenue once granted. The company was founded by the brothers Alexander and Andreas Stroe, went through Y Combinator, and announced a 2.8 million dollar seed round in April 2026 led by FundersClub with participation from Y Combinator, Transpose Platform and Blast.
Capability grades
All 15 axes, graded from public sources on the date shown. Hover a grade to see what the letter means on that axis.
AI Centrality
How much of the product is actually AI. Whether the machine learning is the mechanism the buyer is paying for or a feature layered onto conventional software, and whether the vendor is specific about which is which.
The models are the entire product and the company says so directly. The seed announcement states that the manual work of reading claims, charting evidence, searching prior art and analysing potentially infringing products has historically been the bottleneck, and that large language models and modern search infrastructure are now at a point where that work can be automated. Every published module is a model function: interpreting a claim and decomposing it into elements, locating supporting passages in product documentation, scoring each element and the claim overall, running prior art and invalidity searches, and analysing a portfolio. There is no conventional product underneath and no licensed content asset, since the substrate is public patent data that only the models make usable for this purpose. Remove the models and nothing remains that a buyer would pay for, and the stated comparator is not other software but attorneys charting claims by hand. Verified 2 September 2026.
Citation Accuracy and Hallucination Disclosure
Whether the vendor publishes measured accuracy on citations and assertions, grounds output to primary sources, and says plainly what its system does when it does not know. Legal has a documented public record of fabricated citations reaching filed briefs, so an untested claim of accuracy is not evidence.
Grounding is demonstrated in an interface example and described nowhere. The home page carries a worked claim chart against a named patent number in which each claim element sits beside a quoted passage from product documentation, an analysis and a numeric score, with an overall match score for the claim, so a reader can see that outputs are meant to be traceable to a source passage. That is a demonstration rather than a disclosure: no published material explains what corpus the evidence is drawn from, how a passage is selected, what the scores measure, or how they are calibrated. No accuracy figure is stated. The home page carries three statistics under a heading about what is needed to monetise intellectual property, labelled faster claim charts, hours saved per chart and patent database size, and all three render as zero because they are animated counters that did not execute; the labels establish that figures are published, the values could not be read, and that is recorded as a limit on this reading rather than as an absence. The strongest accuracy claim located is in the seed announcement, which states the work can now be automated reliably and produce production-quality deliverables rather than rough drafts. That is an assertion without measurement. Verified 2 September 2026.
Autonomy and Oversight Model
What the system decides on its own, what a lawyer must approve, and whether the vendor documents where the review point sits. A tool that drafts under review and a tool that files without one are different products and different risks.
Nothing published addresses oversight. Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026, which together comprise the whole public site. No statement describes what the platform completes without a human, where a review point sits, what a user should verify before relying on a claim chart, or what the system does when the evidence for a claim element is weak. There is no terms of service and no acceptable use policy in which such an obligation could be placed. The gap is sharper than the grade alone conveys because of what the outputs are for: the company describes the product as generating claim charts used to identify litigation targets, support cease-and-desist correspondence and find counter-claims, and the seed announcement describes the outputs as production-quality deliverables rather than rough drafts. A platform producing assertions of infringement against named third parties, with no published account of who checks them, sits at the bottom of this axis. Verified 2 September 2026.
Operational and Outcome Evidence
Named, dated evidence that the product works in production at real firms or legal departments. Case studies with figures and identified customers count. Unattributed testimonials and launch announcements do not.
No deployment evidence of any kind was located. Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. There is no customer page, no logo strip, no case study, no testimonial and no named user anywhere on the site, and the news section contains a single item, the April 2026 funding announcement. The three statistics the home page does carry are unattributed and rendered as zero because the counters did not execute, so even the vendor's own headline figures could not be read, and none of the three is a deployment measure in any event. The only outcome language located is a claim that a portfolio can start generating revenue within ten minutes, which appears in vendor-written copy on a third-party startup directory rather than on the company's own site. Nothing published names a customer, a sector, a portfolio size or a result. Verified 2 September 2026.
Privilege and Confidentiality Posture
How client confidences are handled: attorney client privilege and work product treatment, segregation of one client matter from another, whether client data trains any model, and what the vendor commits to in writing rather than in marketing.
Not one of the five limbs is addressed. Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. Nothing states whether customer content is used for training, nothing describes segregation between customers, neither privilege nor work product is mentioned, no retention or deletion position exists, and no model provider is identified. There is no terms of service, no privacy policy, no data processing addendum and no security page on the site in which any of it could appear, and a targeted search returned no such document either. The material a customer would upload makes this consequential rather than merely incomplete: the workflow requires a company's own patent portfolio and, for the analysis to be useful, its assessment of which competitors may be infringing, which is pre-litigation work product a patent owner would ordinarily treat as privileged and would not want visible to anyone else. Verified 2 September 2026.
UPL and Professional Responsibility Posture
Whether the vendor is clear that it supplies a tool rather than legal advice, who its audience is, and how it addresses unauthorized practice of law, competence and supervision duties, and jurisdiction limits. ABA Formal Opinion 512 is the reference point.
Nothing published addresses the advice line. Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. There is no statement that output is not legal advice, nothing on who may use the platform or under what supervision, no reference to verification duties, and no jurisdictional statement, and there is no terms of service or disclaimer page anywhere on the site in which such a statement could sit. The outputs make the absence material. A claim chart is a litigation instrument: it is the document a patent owner uses to allege that a specific product practises each element of a claim, and the company describes its charts as supporting cease-and-desist correspondence and the identification of litigation targets. The product also runs invalidity analysis intended, in the company's own framing, to establish that a patent will hold up in court. A platform generating infringement contentions and invalidity positions, publishing nothing at all about the professional obligations of the person who sends them, sits at the bottom of this axis. Verified 2 September 2026.
AI Governance and Bias Disclosure
Published governance over model behaviour: who owns it inside the vendor, what is tested before release, and what is disclosed about disparate output across matter types, parties, or populations.
No governance material of any kind was located. Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. There is no responsible AI page, no governance framework or set of principles, no individual or function named as accountable for model behaviour, no account of what is evaluated before a model change ships, and no certification. Nothing anywhere addresses uneven output, which on this product would mean whether claim interpretation and evidence matching perform differently across technical fields, patent offices, document languages or the quality of available product documentation, all of which plainly vary and all of which shape which companies get named as infringers. The only statement in the vicinity is the seed announcement's assertion that the underlying technology has reached a point where the work can be automated reliably, which is a claim about the field rather than a description of any control the company applies. Verified 2 September 2026.
AI Safety and Data Stewardship
Retention, deletion, access control, and what happens to prompts and documents after they are processed. Whether the vendor states its subprocessors and its incident practice, or leaves the buyer to assume.
None of the five elements is published. Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026, and ran a targeted search for company security or legal documentation which returned nothing. There is no security page, no privacy policy and no data processing addendum. Nothing states any retention period or deletion practice, no subprocessor is named, no incident or breach notification practice is described, no encryption or access control is mentioned, and no trust centre exists. A login exists at the platform subdomain, so accounts and stored customer data plainly exist, and nothing published describes how either is protected. Under the rule that a site's own navigation establishes its page inventory, this is a thin site rather than an unreachable one: every page rendered cleanly and the whole public estate is a home page, a one-item news index and one article. Verified 2 September 2026.
AI Liability and Recourse
What the vendor stands behind contractually when its output is wrong. Indemnities, caps, carve outs, insurance, and whether any of it is published or only reachable through a negotiated agreement.
No agreement of any kind is published, so nothing addresses who bears the loss when the output is wrong. Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026: the footer carries only a contact email address and a LinkedIn link, with no legal section, no terms of service, no privacy policy and no acceptable use policy, and a targeted search returned no such document from the company. There is no liability cap, no indemnity in either direction, no warranty, no service level and no insurance position on any reachable surface. The exposure this product creates is unusually concrete and unallocated: a claim chart asserting that a named company infringes is the basis on which a cease-and-desist letter is sent, and an error in it can expose the sender rather than the vendor. Nothing published says who carries that. Verified 2 September 2026.
Practice Systems Integration Depth
How deeply the product reaches into the systems legal work already lives in: document management such as iManage and NetDocuments, Word and Outlook, contract lifecycle management, matter management, e-billing, and court filing systems.
No integration into the systems legal or intellectual property work already lives in was located, and none is claimed. Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. No intellectual property management, docketing, document management or matter management system is named, no connector or integration is described, no API or developer documentation exists, and there is no integrations page in the navigation. The published workflow is self-contained: a user uploads patents and receives claim charts through the platform's own interface, reached by a login at the platform subdomain. Nothing describes export formats either, which matters because a claim chart's purpose is to be attached to correspondence or a filing rather than viewed in a dashboard. Verified 2 September 2026.
Deployment Model and Data Residency
Where the software runs and where the data sits. Multi tenant cloud, single tenant, private deployment, on premises, and whether region of residence is a published option or an enterprise conversation.
Nothing published addresses where the platform runs or how customers are separated within it. Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. No cloud provider is named, no hosting region or country is stated, no residency option is offered, nothing distinguishes processing from storage, and no tenancy or isolation model is described. Cloud delivery is evident only by inference from a browser login at a platform subdomain, and even that is not stated as a deployment model. There is no security page or privacy policy in which any of it could appear. The company is described in third-party records with addresses in more than one country, and no first-party source states an operating location, so nothing published would let a buyer with data residency obligations assess the product at all. Verified 2 September 2026.
Security Certifications and Trust Center
Independent attestation a buyer can pull without a sales call: SOC 2, ISO 27001, penetration test summaries, a trust center with current reports and named scope rather than a badge image.
No certification is claimed and no trust surface exists. Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026, and ran a targeted search for company security documentation which returned nothing. Neither SOC 2 nor ISO 27001 nor any other standard is named anywhere, no auditor or attestation is referenced, there is no trust centre or portal, no badge or seal appears on any page, and no penetration testing or vulnerability programme is described. This is an absence rather than a retrieval limit: every page on the site rendered cleanly and the navigation establishes the full page inventory, so there is no gated or unreachable security surface behind which such material might sit. Verified 2 September 2026.
Model Supply Chain Disclosure
Which models sit underneath, whose they are, where they run, and whether the vendor commits to telling customers when that changes. A legal buyer inherits every dependency it cannot see.
Nothing identifies what powers the product. Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. No model is named, no version, no provider entity, and nothing states whether any component is built in-house or reached through a third party. Nothing describes where inference runs, what any provider may retain, or whether customers would be told if the model set changed. The only reference to the underlying technology is in the seed announcement, which observes that large language models and modern search infrastructure have reached a point where this work can be automated; that is a statement about the state of the field rather than a disclosure of this company's stack, and it names nothing. There is no subprocessor list, no privacy policy and no terms of service in which such a disclosure could appear. Verified 2 September 2026.
Commercial Transparency
Whether a buyer can learn what this costs without entering a sales process: published rates, the unit being charged, what sits behind an enterprise tier, and what implementation adds.
No pricing information is published at any level. Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. There is no pricing page, no tier or plan is named, no unit of charge is identified, and no structure is described, whether per user, per patent, per chart, per portfolio or by subscription. Nothing states whether onboarding is charged separately, and no trial or free tier is offered. The only commercial routes on the site are a Book a Demo control in the header and at the foot of the home page, and a contact email address. The one adjacent figure published concerns the cost of the problem rather than the price of the product, with the company noting that obtaining and maintaining a single patent can cost around fifty thousand dollars over its lifetime. No pricing row is written, because the rule that triggers one requires published structure and none exists. Verified 2 September 2026.
Firm and Practice Coverage
Who the product is actually built for. AmLaw, midlaw, small firm and solo, in house departments, government and courts, and which practice areas are supported rather than merely claimed.
Neither the buyer nor the boundary is stated. Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. There is no page addressed to any buyer segment: nothing distinguishes in-house intellectual property teams from law firms, licensing companies, universities or individual inventors, and no industry or technology field is named as supported. No jurisdiction or patent office is identified anywhere, so nothing states whether the claim charting and prior art work covers United States practice only or reaches the European Patent Office and other offices, which is the first question a portfolio holder filing internationally would ask. No portfolio size band is given and nothing states what the platform does not cover. What is published is a set of use cases rather than a coverage statement, running to licensing and litigation targeting, cease-and-desist support, counter-claim discovery, valuation in wind-downs and diligence, and most of that list appears in vendor-written copy on a third-party directory rather than on the company's own pages. Verified 2 September 2026.
Legal Signals
What each signal meansA signal records what public sources say on the date shown. It is not a grade and it is not a recommendation. Where a signal reads Not addressed, it means the index did not locate the material in public sources on that date, which is a statement about disclosure rather than about the product.
Client Data in Training
Can material a lawyer puts into this product be used to train a model?
No located term or policy addresses the question either way.
Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026, and ran a targeted search for company legal documentation which returned nothing. Nothing addresses whether customer content is used to train or improve models, in either direction. No clause names training, model training, machine learning or model improvement in relation to customer content, and there is no terms of service, privacy policy or data processing addendum on the site in which such a term could sit. Nothing on any surface promises the opposite either. The silence has weight on this product because the material a customer uploads is its own patent portfolio together with, in use, its working view of which competitors may be infringing, and because vendor-written copy elsewhere describes a proprietary database of patent and product relationships that grows with each client, which is the shape of an accumulating dataset without any statement of what feeds it or what is done with it.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
No located public material states how long prompts and outputs are retained.
Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. Nothing addresses retention of any kind. No period is stated for uploaded patents, generated claim charts, prior art results or portfolio analyses, nothing describes what happens to that material when a customer stops using the platform, and no deletion route, export right or retention setting is published. The reason is structural rather than an omission from one document: there is no terms of service, no privacy policy and no data processing addendum anywhere on the site, so the documents in which a retention position normally lives do not exist, and a targeted search returned no such document either. A login exists at the platform subdomain, so uploaded portfolios and generated charts are plainly stored somewhere, and nothing published says for how long.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
No located public material addresses walls or matter level segregation.
Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. Nothing describes segregation between customers, isolation of one portfolio from another, or any access boundary. No tenancy model is stated and no permissions or role structure is described. The question has a specific edge on this product that nothing published reaches. The platform's value depends on mapping patents against competing products across a market, so two companies competing in the same technical field could plausibly both be customers, each having uploaded its portfolio and each being a potential target in the other's analysis. Nothing published states what separates them, and vendor-written copy elsewhere describes a proprietary database of patent and product relationships that expands with each client, which raises the question rather than answering it.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
No located term or policy addresses third party requests for customer data.
Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026, and ran a targeted search for company legal documentation which returned nothing. Nothing addresses disclosure to authorities or in response to legal process, and nothing addresses notice to the customer if a demand arrives. The question is not reached rather than answered adversely, and the reason is that no terms of service, privacy policy or confidentiality provision is published anywhere. The gap is worth naming on this product because the stored material is pre-litigation analysis: a customer's assessment of which third parties may be infringing its patents is exactly the material an opposing party would seek in discovery once enforcement begins, and nothing published states what the vendor would do if asked for it.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
No located public material identifies the corpus behind the product’s answers.
Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. The corpus is referred to and never identified. The home page carries a statistic labelled patent database with a figure in millions, but the counter renders as zero because it did not execute, so the scale claim exists and could not be read. No patent office, database, publisher or data supplier is named, no jurisdictions are listed, and no update cadence is stated. Two other bodies of material are plainly in use and are described nowhere: the product documentation and specifications from which evidence passages are drawn for claim charts, since the worked example on the home page quotes product literature directly, and the prior art sources behind the invalidity analysis. No licence, agreement or permission basis is stated for any of it.
Good Law Verification
Does the product tell you when the authority it just cited has been overruled?
No located public material addresses whether authority is checked for subsequent history.
Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. Nothing addresses whether material surfaced by the platform remains current or valid. The product does not retrieve primary law, so a citator has nothing to operate on and the limb does not bite in its usual form. The analogous question is live here and unaddressed: nothing states whether the platform checks that a patent it charts is still in force, whether maintenance fees have been paid, whether claims have been amended or cancelled in post-grant proceedings, or whether a competing product it identifies is still on the market. A claim chart built on a lapsed or amended patent is worse than no chart, and nothing published describes a status check of any kind.
Refusal and Uncertainty Behaviour
What does the product do when the answer is not in the corpus?
The product exposes a confidence or grounding score without an explicit abstention path.
Confidence is scored and surfaced, and nothing describes what happens at the low end. The worked claim chart on the home page shows a numeric score against each claim element, running from 95 down to 45 across the four elements displayed, alongside an overall match score of 85 out of 100 for the claim. Scoring each element separately rather than returning a single verdict is a meaningful design choice, because it shows a user which limb of a claim the evidence is weakest on, and the example deliberately includes a low-scoring element rather than a clean match. What is absent is everything around it. Nothing states what the scores measure or how they are calibrated, no threshold is published below which a match is not asserted, and nothing describes whether the system declines to chart a claim it cannot support or simply returns low numbers. There is no documentation surface on the site in which any of that could appear.
Fabricated Citation Record
Does a public court record exist involving output from this product?
No court order, opinion or disciplinary record naming this product has been located as of the date shown. This is a statement about the public record, not a finding about the product.
Searched the AI Hallucination Cases database maintained by Damien Charlotin, and reporting drawing on it, on 2 September 2026 on the product names PatentWatch and Patent Watch. No court order, opinion or disciplinary record naming the product was located. This is a statement about the public record rather than a finding about the product, and two things bound it. The company is roughly a year old and its user base is not disclosed, so there has been little time and little visible surface for an incident to arise or be traced. And the failure mode this product would produce is not the one the tracker records: an unsupported infringement contention or a missed piece of prior art surfaces as a weak case, an invalidity ruling or a fee-shifting motion rather than as a sanctions order about fabricated citations, and no tracker indexes those.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
No located public material engages with bar or ethics guidance.
Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. No public material engages guidance from any professional body. Neither the USPTO Rules of Professional Conduct nor its guidance on artificial intelligence in submissions is named, no bar opinion is cited, and nothing addresses the obligations of a practitioner relying on a generated claim chart or invalidity position. Nothing addresses Rule 11 or the pre-suit investigation duty either, which is the provision that bites first on this product, since a claim chart is the usual evidence that a patent owner conducted a reasonable pre-filing inquiry before asserting infringement. There is no terms of service or disclaimer page in which any such reference could sit.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
Public materials claim time savings without addressing billing or disclosure.
Time savings are claimed and the figures could not be read. The home page publishes three statistics under a heading about what is needed to monetise intellectual property, two of which are savings measures labelled faster claim charts and hours saved per chart. Both render as zero because they are animated counters that did not execute, so the claims are established and their values are not, and that is recorded as a limit on this reading. Nothing addresses how AI-assisted work is recorded, billed or disclosed to a client, and no per matter record of AI-assisted work was located. The question is live rather than academic on this product: claim charting is work that outside counsel bills for, often substantially, and the company positions itself explicitly against attorneys doing that work manually over weeks, so a firm using the platform faces exactly the compression this signal exists to track. Nothing published engages it.
Outside Counsel Guideline Readiness
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
No located public material supports a client side disclosure obligation.
Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026, and ran a targeted search for company legal or security documentation which returned nothing. No subprocessor list exists in any form, no processor or vendor is named anywhere, and no model provider is identified. There is no data processing agreement, no privacy policy, no terms of service and no security page, so no forwardable client-facing artifact of any kind exists. There is no trust centre. A firm asked by a client which entities would process its patent portfolio and its infringement analysis could answer nothing from anything the company publishes, and would have to obtain every element of that answer by direct request.
Court Disclosure Support
If a judge’s standing order requires an AI disclosure, can the product produce one?
No located public material addresses court disclosure or verification certification.
Checked the home page in full, the news index, the seed announcement in full and the site navigation and footer on 2 September 2026. Nothing addresses disclosure of AI use to a court or patent office, and no exportable record of which model produced which determination, on what confidence, or who reviewed it is described. The product does generate the artifact that would need to carry such a record: a claim chart is filed, served or attached to correspondence, and the home page example shows each element with its supporting passage and score, so the traceability material exists inside the output. Nothing states that it is exportable, nothing describes a review or approval step, and nothing addresses what a party should say if asked how the chart was produced. The gap is pointed here because a claim chart is often the document offered to show that a reasonable pre-suit investigation took place.