PatentPal
PatentPal automates the mechanical writing in a patent application. A drafter supplies the claims and the software generates the supporting specification language from them, together with the figures that go with it, covering flowcharts for methods and block diagrams for systems and devices, the detailed description of each figure, and the abstract and summary. Drafts export to Word and to Visio or PowerPoint, and generation settings can be saved as profiles and switched between so that output follows a particular drafter's or firm's preferred phrasing. The intended user is the patent attorney or agent who already has claims in hand, and the company's stated aim is to return the hours otherwise spent transcribing claims into specification prose. PatentPal Inc. was founded in Silicon Valley by Harvard alumni combining patent law and machine learning backgrounds, is led by founder and chief executive Jack Xu, a former patent attorney, and has raised a financing round led by Foothill Ventures. Its published terms state that patent documents are processed in memory on Amazon Web Services and not otherwise stored, that their contents and any generated output are not shared with any third party other than AWS, and that all data processing takes place within the United States. The terms also state expressly that using the service creates no attorney-client relationship and that its output is not legal advice.
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 generation model is the entire product. PatentPal takes a set of claims and produces the specification language, the figures, the detailed description of each figure and the abstract, and there is nothing else in the offering: no repository, no search, no docketing, no matter management. The founder's own account on the About page describes the origin as writing the code for an initial summary generator and watching pages of writing appear at the push of a button. Remove the model and there is no product to sell, not a diminished one. Checked 4 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.
No accuracy content of any kind is published, and the only statement on the subject is a disclaimer. The terms of service state that PatentPal makes no guarantees or warranties on the outcome of a patent application or on the quality of its output including the generated language, and specifically do not warrant that the service will be accurate or reliable or error-free. There is no accuracy figure, no test set, no evaluation, no description of how generation is constrained, and hallucination is not addressed anywhere under any name. Nothing describes the relationship between the input claims and the generated specification, which is the grounding question for this product class and the one a drafter would most want answered, since generated support that does not match the claims creates written-description problems. The site inventory is the reason this cannot be softened: the navigation carries only About and an external Medium blog, so **there is no product or technology page on which such material could sit**. The Medium blog was not opened in this pass. Checked 4 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.
Autonomy is the selling point and oversight is asserted only through a legal disclaimer. The published framing is one-click generation, with the founder describing pages of writing produced at the push of a button, and the product's value proposition is that the drafter does not do this work. Against that, the terms place the whole burden on the user by three separate routes: no attorney-client relationship arises, output is not legal advice and is not a substitute for advice from a qualified attorney, and the service is used solely at the user's own risk. None of those describes a mechanism. Nothing published identifies a review point in the workflow, no confidence or uncertainty signal is described, nothing states which parts of a draft a professional must check, and nothing addresses what happens when generated support conflicts with the claims. The profile feature that tailors generated phrasing is a configuration control rather than a review surface.
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 production evidence was located. What the About page publishes is investor and competition standing, not deployment: a financing round led by Foothill Ventures, first place among 190 Harvard-affiliated startups at the Harvard China Forum, top five among 43 at the Harvard New Venture Competition, and first place among 15 startups at Plug and Play. Those are credible third-party validations of the company and say nothing about how the product performs in a practice. No customer is named anywhere, no logo strip appears, no testimonial is published, no case study exists and no figure for time or cost saved is attached to anyone. One clause implies customers exist without identifying any: the privacy policy reserves the right to publicise that a user's company is a PatentPal customer, with an opt-out on request. The Medium blog was not opened in this pass and is the only remaining surface that could carry deployment material.
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.
Two substantive commitments, both aimed squarely at the risk this product actually creates, and three limbs absent. The risk is that an unfiled invention disclosed to a third party can destroy novelty, and the terms address it directly on both fronts. Storage: the privacy policy states that patent documents are processed in memory and not otherwise stored, and neither the documents nor any content generated from them is retained except as necessary to provide the service. Third parties: both documents state that the contents of patent documents and any generated output are not shared with any third party other than for processing on Amazon Web Services, which is a direct answer to what an underlying provider may see and a stronger one than most subprocessor lists give. Processing is stated to occur wholly within the United States and to travel over HTTPS. What is missing holds it at B. Nothing addresses training on customer content in any form. Nothing addresses privilege or attorney work product, on a product handling unfiled applications. And segregation is covered only by a prohibition on sharing an account, with no role or permission model described.
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.
A real position is published on advice versus tooling, in a dedicated clause rather than buried in a warranty section. The terms carry a heading of its own, No Attorney-Client Relationship or Legal Advice, and state that the services are not provided as part of an attorney-client relationship, that using them does not give rise to one, and that information received does not constitute legal advice and is not a substitute for obtaining legal advice from a qualified attorney. That covers both limbs a patent drafting tool needs, since the product generates text that goes into a filed instrument. What is missing from the top band is the supervision and competence dimension and any statement of scope: nothing describes what a professional must check before filing, no rule of professional conduct or duty of candour to the patent office is named, no jurisdiction limit is stated although the product generates United States-style specification language, and nothing states who may use the service beyond a representation of authority to bind an entity.
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 position is published for a system that writes text into filed patent applications. There is no responsible AI statement, no governance framework, no named owner accountable for model behaviour, no description of evaluation or testing before release, no published results and nothing on bias. The site inventory makes the absence structural rather than an oversight of placement: the navigation carries About and an external blog only, and the two legal documents are the entire published estate, both dated 1 August 2022. Nothing in either addresses how the model is developed, monitored or corrected. Searched the About page, the terms of service and the privacy policy on 4 September 2026; the Medium blog was not opened.
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.
Two limbs are genuinely well answered and the rest of the set is absent. Retention is addressed in specific technical terms, with patent documents processed in memory and not otherwise stored beyond what is necessary to provide the service, which for this product class is a stronger position than a stated period would be. Subprocessors are named rather than gestured at, and named by data type, which is unusual: account and website information may be shared with Amazon Web Services, Google, LinkedIn and Stripe, while the contents of patent documents go to no third party other than AWS, and payment data is handled by Stripe rather than stored. What is missing is most of the rest. Access control is described only as HTTPS in transit, with no encryption at rest stated, no multi-factor authentication, no role model and no logging. No incident or breach notification practice was located anywhere. There is no security page in the site inventory at all. Both documents carry an effective date of 1 August 2022, four years before this check, and the site footer still reads copyright 2022.
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.
Liability is addressed only through a limitation clause, and the clause goes further than the usual one. The terms disclaim all warranties, expressly state that no guarantee or warranty is given on the outcome of a patent application or on the quality of the generated language, and record that the service is used solely at the user's own risk. The limitation itself excludes not only indirect, incidental, special, consequential and exemplary damages but **direct damages as well**, and does so without stating any cap, so there is no figure a buyer can point to as the floor of their recourse. No indemnity runs in either direction, and no insurance position was located. What keeps this off the bottom of the band is that the allocation is published, short, and readable in full before signing, and that the exposure the product creates is named rather than left implicit: the terms address the outcome of the patent application specifically, which is the loss a drafter actually fears.
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 practice systems was located on any first-party surface. There is no integrations page, no API reference and no developer documentation, and the site inventory is exhausted by About, an external Medium blog, a login and two legal documents. No document management system, IP management system, docketing platform or patent office filing interface is named anywhere. Export to Word and to Visio or PowerPoint is described consistently across third-party reviews and is plainly a real capability, but it appears on no first-party surface read and is in any case export rather than integration. The product is used through a separate application at draft.patentpal.com, and nothing describes how drafts move between it and the systems a firm already runs. The Medium blog was not opened in this pass and is named so the limit is visible. Checked 4 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.
Residency is stated plainly and tenancy is not addressed. The terms record that all data processing is done within the United States, and both the terms and the privacy policy identify Amazon Web Services as the environment in which patent documents are processed, with data sent over HTTPS. A country and a named infrastructure provider is a real residency position rather than a region label, and it is the limb that matters most to a buyer worried about where an unfiled invention travels. What is absent is the rest: nothing states whether the platform is single or multi-tenant, no dedicated or private option is offered, no AWS region is named, and nothing distinguishes where processing happens from where anything is stored, although the in-memory processing statement makes the storage question largely moot by design.
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 independent security attestation was located, and no certification is claimed. There is no SOC 2, ISO or other framework named anywhere, no auditor, no report at any access tier including on request, no penetration testing mentioned, and no trust centre or security page: the site navigation carries About and an external blog, and the footer carries the two legal documents and an email address. Because nothing is claimed, there is also nothing unsupported on display, which is why this rests on absence rather than on overclaiming. The only security-adjacent third-party evidence points at the host rather than the vendor: the terms direct the reader to Amazon's service terms, customer agreement and privacy notice and state that Amazon is similarly committed to data privacy, which is AWS's posture and does not travel to PatentPal. Checked About, terms, privacy policy and footer on 4 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.
The vendor answers who does not touch the data and never says what does. Both published documents state that the contents of patent documents and any generated output are not shared with any third party other than for processing on Amazon Web Services, which read against the in-memory processing statement implies that generation runs on models PatentPal controls within its own AWS environment. That is a real negative disclosure and it is the half a patent drafter cares about most. The positive half is entirely absent: no model is named, no version or family is given, no third-party model or inference provider is identified, no AWS region or service is specified, and no commitment to notify customers if the model or its hosting changes was located. Naming the cloud says where the software runs, not whose model writes the specification. Both documents are dated 1 August 2022, which predates the current generation of language models, so even the implication rests on a four-year-old statement.
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, including the unit of charge. There is no pricing page in the site inventory: the navigation carries About and an external Medium blog, and the footer carries the terms, the privacy policy and an email address. Nothing states whether charging is per seat, per application, per generation or per subscription period, and no figure, band or range appears anywhere. What the terms do establish is only the shape of the commercial relationship, not its cost: payment is required before the services start, for example at the beginning of each subscription period, all payments are non-refundable, and prices and payment terms may be updated at any time with changes applying prospectively. A trial and a price point are described in third-party software directories and reviews; those are not first-party and are not credited. No VendorPricing row is written, because nothing lifted this axis off the floor. Checked 4 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.
No coverage documentation was located. Nothing published describes who the product serves, what technology fields it handles, which jurisdictions its output is written for, or where it stops. The buyer is identifiable only by inference from the About page, which says the company builds software to auto-generate language for patent applications and describes the founder's background as a patent attorney, and from the terms, which contemplate agreement by a law firm, company or other entity. No segment page, no practice page, no firm size, no in-house or government treatment, and no statement of limits exists. The absence is structural rather than a matter of placement: with no product page anywhere in the site inventory, there is nowhere for coverage material to sit. This is one of the thinnest published estates in the corpus, and R20 applies: a thin site is a finding about the vendor rather than a limit on the researcher, since every page that exists was reached and read.
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.
No located term or policy addresses training either way. Both published documents were read in full on 4 September 2026 and neither uses the word or engages the concept. What they do address is storage and sharing, and tightly: patent documents are processed in memory and not otherwise stored except as necessary to provide the service, and their contents and any generated output are not shared with any third party other than for processing on Amazon Web Services. Those constrain what happens to a document without stating whether it contributes to model improvement, and a document processed in memory can still inform a model during that processing. Recorded as silent rather than as a commitment, because neither document says what a buyer would need it to say. Searched the About page, the terms of service, the privacy policy and the site footer; there is no security or product page in the site inventory.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
Retention is acknowledged in public materials with no stated period.
Retention is addressed in technical rather than temporal terms, and no period is stated because the position is that nothing is kept. The privacy policy states that patent documents are processed in memory and not otherwise stored, and that neither the documents nor any content generated from them is retained except as necessary to provide the software services; the terms of service repeat the same commitment. For a tool handling unfiled applications that is a stronger answer than a stated window would be. What keeps this from the top of the scale is the qualifier: except as necessary to provide our services is undefined, nothing states how long that necessity lasts for a draft in progress, no deletion route or confirmation is offered, and zero retention is not presented as a customer-selectable setting. Both documents carry an effective date of 1 August 2022.
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.
No located public material addresses walls or matter level segregation. The only access provision is a prohibition in the acceptable uses section of the terms on sharing an account with any other person or entity, which allocates responsibility rather than describing separation. Nothing states whether the platform is single or multi-tenant, no role or permission model is described, no administrator function is mentioned, and nothing addresses how one drafter's work is kept from another's inside a firm account. The question has real weight for this product class because a patent firm routinely drafts for competing clients in the same technology field. The in-memory processing statement reduces the exposure without answering the question. Searched the About page, the terms and the privacy policy on 4 September 2026.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
Published terms or policy address disclosure to authorities or in response to legal process, and no commitment or reservation regarding customer notice is located anywhere. The vendor has told the customer that data can leave and has said nothing about whether the customer hears of it.
The privacy policy addresses disclosure under legal compulsion and commits to no notice. It provides that PatentPal may share information it collects to comply with legal obligations and to enforce its terms of service, with no notice commitment to the customer and no discretion over notice reserved. One tension is recorded rather than resolved, because it cuts in the customer's favour: the same policy states separately that the contents of patent documents and any generated output are not stored or shared with any third party other than for processing on Amazon Web Services, and that sentence carries no legal-process carve-out on its face. So the compelled-disclosure clause plainly reaches account and website information, and its application to the patent documents themselves is left unresolved by the drafting. No transparency report was located.
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.
One sentence touches the question and identifies nothing. The privacy policy states that PatentPal may collect publicly available patent data and other information from public sources, which acknowledges that a corpus beyond the user's own claims exists without naming any part of it. No database, patent office feed, publisher or aggregator is identified, no jurisdictional coverage is described, no update cadence is given, and no licensing basis is stated for any of it. The product's primary input is the user's own claims, from which the specification is generated, so the corpus question bites less hard here than on a search product, but the public-sources sentence shows the vendor knows the question exists and leaves it unanswered. Searched the About page, the terms and the privacy policy on 4 September 2026.
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.
Nothing on any located surface addresses whether authority is checked for subsequent history. The product does not retrieve or cite prior art, case law or patent status: it generates specification language, figures and an abstract from claims the drafter supplies. The question does not bite on this product class and the honest value is the absence rather than a penalty. Searched the About page, the terms and the privacy policy on 4 September 2026.
Refusal and Uncertainty Behaviour
What does the product do when the answer is not in the corpus?
No located public material addresses what the product does when it cannot ground an answer.
No located public material describes what the product does when it cannot generate reliably. Nothing describes an abstention path, a no-answer state, a confidence or coverage signal against generated text, or any flag where a claim element lacks adequate support in the generated description. The published framing runs the other way, presenting one-click generation of a complete draft, and the terms handle the risk by disclaimer rather than by behaviour, stating that no warranty is given that the service will be accurate or reliable or error-free and that use is solely at the user's own risk. Searched the About page, the terms and the privacy policy on 4 September 2026; there is no product page in the site inventory and the Medium blog was not opened.
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.
The AI Hallucination Cases database maintained by Damien Charlotin was searched on 4 September 2026 on both the product name PatentPal and the corporate name PatentPal Inc. No court order, opinion or disciplinary record naming the product was located. The database tracks fabricated legal citations in court filings; this product generates specification prose from claims rather than citing authority, so its exposure to that specific failure is structurally low, though generated support that does not match the claims is a distinct risk the database does not capture. This records the state of the public record on that date and is not a finding about the product.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
Public materials refer to professional responsibility in general terms without naming guidance.
Public materials engage professional responsibility in general terms without naming any authority. The terms carry a dedicated clause stating that no attorney-client relationship arises from use of the service and that its output does not constitute legal advice and is not a substitute for advice from a qualified attorney, which addresses the subject squarely. No bar association, patent bar, regulator, rule of professional conduct or ethics opinion is named anywhere, and nothing addresses the duty of candour to the patent office or the drafter's own obligations when filing generated text over their signature, which are the professional duties this product most directly engages. Searched the About page, the terms and the privacy policy on 4 September 2026.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
No located public material addresses billing, fee or disclosure treatment.
No located public material addresses billing or fee disclosure, and unusually there is not even a savings claim on the record to record. The About page describes the aim as giving valuable time back to the attorney so it can be directed towards maximising client value, which is a purpose statement rather than a quantified saving, and no figure for time or cost saved appears anywhere. Nothing addresses what happens to a fixed drafting fee or a billed hour when specification writing is automated, and no per-matter record of AI-assisted work is described. The gap is notable because patent drafting is frequently billed at a flat fee per application, so the compression this signal was written for lands directly on this product's economics. Searched the About page, the terms and the privacy policy on 4 September 2026.
Outside Counsel Guideline Readiness
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
A current subprocessor or model provider list is published.
Some of the material a firm would forward exists, published rather than gated, and the model layer is missing. The privacy policy names the third parties that receive information and does so by data type, which is better than most: account and website information may go to Amazon Web Services, Google, LinkedIn and Stripe, while the contents of patent documents and any generated output go to no third party other than AWS. Both documents are public and could be sent to a client as they stand. What is absent is the disclosure a client's AI clause actually asks for: no model provider is identified anywhere, no model is named, there is no data processing agreement, no consent or notification pack, and no security documentation at any access tier. Naming the cloud and the marketing vendors does not tell a client whose model saw the invention.
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.
No located public material addresses producing a record of AI-assisted work. Nothing describes an export, log or audit trail recording which text was generated, which model produced it, what the drafter changed before filing, or when any of it happened. The in-memory processing position makes such a record structurally unlikely, since nothing is retained to export from. The question is live for this product class rather than academic: generated specification text is filed over a practitioner's signature at a patent office, and a firm may need to account for how a description was prepared. Nothing published engages that, and no disclosure guidance or template was located. Searched the About page, the terms and the privacy policy on 4 September 2026.