T
Triangle IP
Triangle IP is a patent pipeline and prosecution analytics platform from Triangle IP, Inc., a Colorado company. Its TIP Tool™ is a Jira-inspired workspace where a company captures invention disclosures, scores and ranks ideas for value and patentability, moves them through drag-and-drop stages with inventors and outside counsel collaborating in the same place, and then tracks the resulting applications as a portfolio with automatic USPTO status updates, family trees, Gantt-style prosecution timelines, tagging by product, deadline reminders and spreadsheet export.
The machine learning sits in the Predictor: it reads an invention summary, a draft or an uploaded file, compares the claim language against public USPTO prosecution data, and returns a ranked list of likely Group Art Units and CPC codes with confidence levels, each mapped against its historical allowance rate, so a drafter can amend claims and rerun the prediction to steer away from a hard art unit. That feeds a patentability score, and sits alongside statistical analytics on examiner behaviour, expected rounds of office actions, prosecution cost and the performance of the firm's own outside counsel.
The Predictor is sold standalone at five dollars a month as well as inside the platform, which is priced flat per organisation with a free tier that never expires. The vendor states that text submitted to the Predictor is discarded once the prediction is generated.
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 drive one real capability inside a workflow product that would run without them. Strip the machine learning out and the TIP Tool is still a working system: a Jira-inspired board where invention disclosures are captured, scored, merged and moved through stages, with inventors and outside counsel collaborating in one place, portfolio views, family trees, Gantt-style prosecution timelines, automatic USPTO status updates, tagging by product and spreadsheet export.
What the models do is the Predictor: read an invention summary, a draft or an uploaded file, analyse the claim language against public USPTO prosecution data and return ranked Group Art Units and CPC codes. That is not a bolt-on either. It is sold as a product in its own right at five dollars a month, it sits in the main navigation, and it feeds the patentability score the platform uses to rank ideas. The examiner and counsel analytics beside it are statistics drawn from USPTO records rather than model output. Verified 20 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.
The predictor is never validated in public, and what it does show the user is better than most. Each CPC prediction comes back in a ranked list with a confidence level, described as showing how strongly the model links the submitted text to that code, and every predicted code is plotted against its historical allowance rate so the reader can see what the prediction implies. The source is named: models trained on the USPTO's public prosecution data.
What is absent is any measurement of whether the predictions are right. No accuracy figure, no test set, no error rate, and nothing on how often a predicted art unit turns out to be the assigned one. The nearest things are directional: a statement that a more detailed input tends to produce a more reliable prediction, and a case study claiming a 30 per cent rise in allowance rate at an unnamed company. The published terms disclaim the accuracy, completeness and usefulness of everything on the estate. Verified 20 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.
The product advises and the human decides, and almost none of that is written down. In practice the machine does not act: it returns predicted art units, CPC codes, confidence levels and allowance rates, and a person chooses whether to amend the claims. But nothing published describes that division. There is no statement of what the model decides on its own, no threshold at which a prediction should not be relied on, no described review point between a prediction and a drafting decision, and nothing on what happens when a prediction turns out wrong.
Access controls are real and documented, a two-tier owner and portfolio-admin model with per-portfolio permissions, but those govern who sees a case rather than who checks the model. The closest published limit is a scope statement about a different feature, that the tool is not a substitute for an attorney's formal docketing system. Verified 20 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.
One figure, one anonymous customer, no method. The estate carries a case study headed as a 30 per cent increase in patent allowance rates for a leading global tech company. The account is coherent, describing a firm that could not tell which innovations to back, used examiner allowance rates and typical rounds of prosecution to forecast cost and difficulty, and questioned whether its existing counsel had the right record in the relevant art unit.
What it does not give is the company, the period, the portfolio size, the baseline, or how the increase was measured. No customer is named anywhere else on the estate either: there are no logos, no named reference accounts and no dated deployments, and the review evidence the vendor points to sits on a third-party software directory. For a product whose whole proposition is that data beats guesswork, the absence of its own measured evidence is the thing to notice. Verified 20 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.
One precise commitment, and no agreement a buyer can read it against. The precise part is about the Predictor, and it answers the question a patent team would ask first: invention text entered or uploaded is not stored, retained or reused, it is processed only at the moment of the request to compare against existing records, and once the prediction is generated the data is discarded rather than saved. Around that sit general assurances of enterprise-grade encryption, United States based servers and role-based access, plus portfolio isolation so one division cannot see another's cases.
The problem is where the commitments live. The product terms and privacy policy sit behind the application login and could not be read, so none of this is checkable against anything binding. The agreement that is published governs the website, permits only personal non-commercial use, and treats material posted through its interactive features as non-confidential. Verified 20 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. Where the advice line is not the duty a product raises, the axis is read through the nearest professional duty it does raise: judicial conduct rules and the reviewing duty for products sold only to courts, and the duty to bill for time actually spent for products that draft time entries.
A disclaimer of accuracy stands in for a position on advice. The published Terms of Use say the information on the estate is for general information only, that the vendor does not warrant its accuracy, completeness or usefulness, and that reliance is at the user's own risk. Nothing says the output is not legal advice, nothing addresses an attorney-client relationship, and nothing addresses the supervision question the product actually raises, which is that a prediction about art unit assignment is used to reshape claim language before filing.
The audience is not ambiguous, and that cuts both ways: patent practitioners and enterprise IP teams have their own pages, and so do individual inventors and startups, who are the users least likely to have counsel reading over the result. The estate does publish a guide to hiring a patent attorney, which is practical rather than a statement about the tool's limits. Verified 20 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 position is published about the model at all. Nobody inside the company is named as accountable for it, nothing describes what is tested before a change to the predictor ships, no retraining cadence or data cut-off is stated, and no principles page exists. The bias question is not abstract for this product: the model is trained on historical USPTO assignment and allowance patterns, it is sold on the basis that some art units grant at one per cent and others at ninety-eight, and its explicit purpose is to steer applications toward the favourable ones.
Whether the model's predictions hold evenly across technology areas, applicant sizes or drafting styles, and what happens when historical patterns encode something a user would not want to optimise toward, is addressed nowhere. Checked the Predictor page, the FAQ, the help centre, the Terms of Use and the full footer on 20 September 2026. Verified 20 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.
The one thing addressed specifically is what happens to submitted text, and the rest is generic. Published: the Predictor discards invention text once a prediction is generated and never retains or reuses it; enterprise-grade encryption, United States based servers and role-based access control; a two-tier admin model with portfolio-level administrators; secure document exchange inside the platform to keep drafts out of email.
What is missing is most of the axis. No retention period for anything held in the platform, no subprocessor list, no incident response or breach notification practice, and no access to the privacy policy that governs the product, which sits behind the login. Deletion runs the wrong way and the vendor says so plainly: patent applications cannot be permanently deleted, only archived, and archived records remain stored in the system with their linked ideas, tags and analytics intact. Verified 20 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.
Liability is handled by a website disclaimer, and the document that would govern the product cannot be read. The published Terms of Use exclude damages of any kind under any legal theory arising from use of or inability to use the site, direct and indirect alike, with an exception only for the vendor's gross negligence or wilful misconduct, and everything is supplied as is with all warranties disclaimed. Indemnity runs one way, from the user to the company.
Arbitration is available at the company's sole discretion, which is unusual in being optional for one side only, under Colorado law and in Colorado courts. Nothing anywhere addresses the loss this product could actually cause, which is claim language reshaped toward an art unit on a prediction that proves wrong. No indemnity, warranty, service credit or insurance is offered, and no service level commitment is published. Verified 20 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.
Real mechanisms, described honestly, and not one named system. What moves is stated: a law firm or docketing system can send weekly asset reports or data feeds which are emailed in and processed, the vendor's team assists with manual imports at onboarding, portfolios export to spreadsheet, and the platform pulls USPTO status weekly by default or daily where outside counsel grants access to a dedicated USPTO customer number through an attorney sponsorship arrangement.
Single sign-on is supported and deep links let a case be dropped into an email or a chat message. The vendor is candid that it goes no further, saying it deliberately avoids Outlook, Adobe and DocuSign integrations to move teams away from email-centric workflows. So no document management system, matter management system, billing system or e-signature tool is named as a connection, and no API or developer documentation is published. Verified 20 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.
The region is stated and the tenancy is not. The vendor says data is held on secure cloud hosting on United States based servers, and the published Terms of Use reinforce the geography from the other direction, stating that the owner is based in Colorado and that the site is provided for use only by persons located in the United States. That is a clear answer to where the data sits. What is not published is anything about how customers are separated from one another: no statement of shared or dedicated infrastructure, no single-tenant or private option, no region choice for a buyer with data outside the United States, and no cloud provider named.
Separation is described only inside an account, where portfolios operate as isolated units with their own permissions and one division cannot see another's cases. No processing location is stated for the prediction models as distinct from storage. Verified 20 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 independent security attestation was located and none is claimed. There is no security page and no trust centre in the navigation or the footer, no SOC 2, ISO or penetration test reference on any page including the Terms of Use, and nothing offered on request. The strongest published phrase is enterprise-grade encryption, alongside United States based hosting and role-based access control, which is a description of practice rather than anything a third party has examined.
The product-level privacy policy that might carry more sits behind the application login and could not be read. The absence is recorded rather than inferred: the navigation and footer were run to the bottom and one search for an attestation returned nothing from the estate. This records what is published, not a finding about the vendor's actual security. Verified 20 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 training data is named, which is the unusual part, and the model itself is not. The vendor states that its predictions come from machine learning models trained on the USPTO's public prosecution data, and describes what they act on: claim language compared against public patent records to produce ranked art units and CPC codes with confidence levels. Because the models are the vendor's own rather than a third party's, none of the usual questions about a model provider arise, and nothing on this estate depends on an outside large language model.
What is not published is anything about what sits underneath: no architecture, no model version or identifier, no statement of when the training data was last refreshed or how often the model is retrained, no processing location for inference, and no commitment to tell customers when the model changes beneath a prediction they have relied on. Verified 20 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.
A buyer can price this entirely from the estate. The platform is $50 per month or $500 per year, flat per organisation, with no per-user fees and most paid plans supporting 100 to 200 users. The Predictor is sold separately at $5 a month for anyone who wants only the prediction. A forever-free tier exists and never expires, capped at 25 active cases, against 100 on the premium plan, and a 30-day full-feature trial runs without a credit card.
The standard plan includes three portfolios; organisations needing more divisions move to a custom tier quoted on request, which is the only unpriced element. What implementation adds is stated too: annual subscribers receive complimentary onboarding and data import, including help with foreign cases, and data is preserved when a free account upgrades. The vendor also publishes what it is cheaper than, which is a claim about others and is not used here. Verified 20 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.
Four buyer segments have their own pages and the functional limits are stated with unusual candour. The segments are startups, enterprise, individual inventors and patent practitioners, and the product is built for the corporate side of patent work: inventors, product and engineering teams, in-house IP staff and the outside counsel they invite in. Practice coverage is US patent prosecution, and the boundary is published rather than implied.
International filings can be catalogued, tagged, documented and tracked, but the deep analytics, allowance-rate prediction and examiner behaviour, are US-centric. Cost analysis does not work for foreign filings at all, and the vendor says so directly, calling it an industry-wide gap and adding that it does not oversell capabilities it cannot deliver accurately. What is absent is any sense of firm size or type on the practitioner side, and government or university technology transfer use is not addressed. Verified 20 September 2026.
5 public documents
The public pages on file for Triangle IP, with the recorded signals each one supports and the date it was last read. Open any of them and check the reading against the record.
-
triangleip.com/terms-of-use4 signals
Client Data in Training, Third Party Request and Subpoena Notice, Bar Guidance Alignment and 1 more
Read Sep 20, 2026
-
Prompt and Output Retention, Ethical Walls and Matter Segregation, Good Law Verification and 1 more
Read Sep 20, 2026
-
help.triangleip.com/knowledge-base/how-to-predict-cpc-codes-using-the-predictor-tool-in-tip-tool1 signal
Refusal and Uncertainty Behaviour
Read Sep 20, 2026
-
Primary Law Corpus Provenance
Read Sep 20, 2026
-
Billing and Fee Posture
Read Sep 20, 2026
Predictor $5 per month standalone; TIP Tool $50 per month or $500 per year flat per organisationUSD, as published, never converted
- Triangle IP publishes its prices, and they are simple.
- The TIP Tool is $50 per month, or $500 per year, charged flat for the whole organisation. There are no per-user fees, and most paid plans support 100 to 200 users.
- The Art Unit Predictor can be bought on its own for $5 per month if that is all you want.
- There is a free plan that never expires, capped at 25 active cases. The paid plan raises that to 100 active cases. Archived and deleted cases do not count toward the cap.
- There is also a 30-day trial with every feature turned on, and no credit card needed. Whatever you put in during the free plan or the trial stays when you upgrade.
- The standard plan gives you three portfolios, which is how the tool separates divisions or business units. If you need more, that is a custom tier and you have to ask for a quote. That is the only price not published.
- Annual subscribers get onboarding and data import thrown in, including help loading foreign cases.
Published ungated, flat per organisation rather than per seat. **TIP Tool: $50 per month or $500 per year**, stated as flat-rate per organisation with no per-user fees; the vendor notes most paid plans support 100 to 200 users so engineering, product, legal and finance can all be in the tool without licensing friction. **Art Unit Predictor standalone: $5 per month**, available separately from the platform or integrated with it. **Forever-free tier**, which never expires and supports a limited number of users and filings, capped at **25 active cases**; the **premium plan raises the cap to 100 active cases**, counted at organisation level across all portfolios, with archived and deleted cases excluded. **30-day full-feature trial**, no credit card required, with data preserved on upgrade. **Three portfolios on the standard plan**; organisations needing more divisions or business units move to a **custom tier quoted on request**, the only element not published.
Annual subscribers receive complimentary onboarding and data import. Custom feature development is offered under the custom tier.
Implementation: None charged, and onboarding is described. Annual subscribers receive complimentary onboarding and data import, with the vendor's team assisting on manual imports including foreign cases so a portfolio is fully populated at the start. Data entered during the free plan or the 30-day trial is preserved when an account upgrades, so nothing is lost in moving to a paid plan. Support covers onboarding, technical issues and feature requests at no separate charge, and the vendor states that custom feature development is available under a custom tier package rather than as an hourly service. No migration, configuration or training fee appears anywhere on the estate.
Confidentiality and data terms: No Business Associate Agreement is offered or referred to anywhere on the estate, and none would be expected: the platform holds invention disclosures, patent applications and USPTO prosecution data, with no health information in its path. Worth recording for completeness rather than as a gap. The nearest sensitive-data question for this product is a different one, and the vendor does answer it: unpublished invention text put into the Predictor is discarded once the prediction is generated, which is the assurance a patent team would want before uploading a draft.
Note: Every figure read from the vendor's own FAQ page, last updated 19 May 2026, and the Predictor page, modified 20 August 2026; both publish rates ungated. entryPriceUsd is recorded as 5, the lowest paid rate at which the graded capability can be bought, which is the standalone Predictor subscription; the platform rate of $50 a month or $500 a year is carried in the display and the basis so the figure is not read as the price of the TIP Tool. A forever-free tier exists, so under the free-tier rule the numeric and display carry the lowest paid rate and the free plan is described in the basis rather than recorded as zero. The only unpriced element is the custom tier for organisations needing more than three portfolios. The vendor's published comparisons against named enterprise IP platforms are claims about other vendors and are not used.
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 customer agreement, terms of service or equivalent contract is published on any surface located, and no policy page states a position on training. Nothing is granted and nothing is withheld, so a client has no term to hold the firm to. Where a policy page does state a position, the row takes the matching policy value instead and the summary records that no agreement exists.
No agreement that governs the product could be read, so no training position can be established. The document published on the website is a Terms of Use last modified 11 July 2021 which by its own words governs the website, permits use for personal non-commercial purposes only, and names two further documents as part of the agreement: a Privacy Policy, whose location it gives as an address behind the application login, and an End User License Agreement that is not published anywhere.
Neither could be reached, and one search returned nothing. The website terms say nothing about model training. The only related statement the estate makes is about retention rather than training, and it is specific and narrow: invention text put into the Predictor is discarded once a prediction is generated. That is recorded on the retention row. Whether anything held in the wider platform is used to train or improve the models is not addressed on any readable surface.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
A specific retention period is published and the customer cannot change it.
For the model, the window is zero and the vendor says why. The FAQ answers the question a patent team asks first, whether confidential or unpublished material can safely go into the Predictor: the tool does not store, retain or reuse invention text entered or uploaded, content is processed only at the moment of the request to compare against existing records and generate the predicted Group Art Units, and once the prediction is produced the data is discarded and not saved in the system.
That is a definite position rather than a vague one. The rest of the platform runs the other way and a buyer should hold both. Patent applications cannot be permanently deleted, only archived, and the vendor states that archived records and all linked ideas, tags and analytics remain stored. No retention period is published for anything else, and the privacy policy that would cover it sits behind the login.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
The product maintains its own permission model, documented, requiring the firm to keep it aligned.
Separation inside an organisation is described in real terms. Portfolios are the unit: each operates as a completely separate world with its own users, cases and settings, so one division cannot see another's data, and a standard plan carries three of them with custom plans accommodating any number of divisions or business units. Permissions sit in a two-tier model, an owner or super-admin who set up the account and holds rights across every portfolio, and portfolio-level administrators who manage users, cases and settings only within their own.
Roles run down to the individual: inventors can view only applications they are associated with and cannot archive anything outside their assigned access, while managers, admins and prosecutors can. Outside counsel are invited into specific cases rather than given general access. What is not described is separation between customers, or any conflicts mechanism for screening a named person from a matter.
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.
Disclosure is addressed and no notice attaches to it. The published Terms of Use reserve the right to disclose a user's identity or other information to any third party claiming that posted material violates their rights, and state that the vendor will cooperate fully with any law enforcement authority or court order requesting or directing it to disclose the identity or other information of anyone posting material through the site.
The user is asked to waive and hold the company harmless from any claim resulting from action taken in the course of such investigations. Nothing commits to telling the customer that a demand has arrived, to waiting before complying, to narrowing what is produced, or to assisting with a protective order. The agreement that governs the product, where a confidentiality clause would normally carry that commitment, sits behind the application login and could not be read. No transparency report is published.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
Sources are identified without stating the licence or rights basis.
The corpus is named and it is public. The vendor states that its models are trained on the USPTO's public prosecution data, that predictions are produced by comparing submitted claim language against public patent records, and that the analytics beside them, examiner allowance rates, typical rounds of office actions, prosecution difficulty and counsel performance, are drawn from the same source. Portfolio status comes from the USPTO too, refreshed weekly by default or daily where outside counsel grants access to a dedicated customer number.
So a buyer can tell what the predictions rest on, which is more than most records on this signal manage. What is not stated is the licensing basis, which for public prosecution data is unremarkable, nor the vintage: nothing says what period the training data covers, when it was last refreshed, or how often the models are rebuilt against newer USPTO outcomes.
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.
This product cites no authority, so there is nothing for a treatment signal to check. It predicts where an application will be examined and what the historical outcomes look like; it does not state what the law is, quote a rule or reference a decision, and no statute, regulation or case appears in its output. What it does publish is data currency rather than authority currency: USPTO status is refreshed weekly by default, or daily where outside counsel grants access to a dedicated customer number so that updates are triggered by USPTO notification emails.
That tells a user how fresh the portfolio status is. It says nothing about how current the model behind a prediction is, which is the nearer question here and is recorded on the corpus row. Checked the Predictor page, the FAQ, the help centre and the Terms of Use on 20 September 2026.
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.
Uncertainty is shown as a number beside each answer, and nothing describes what happens at the edges. Every CPC prediction comes back in a ranked list with a confidence level, documented as showing how strongly the model links the submitted text to that code, and the predicted codes are plotted against their historical allowance rates so a user can see both how sure the model is and what the answer would mean. The art unit prediction works the same way and the vendor notes that a fuller input, claims or a strong abstract, tends to produce a more reliable result.
What is not published is any behaviour at low confidence: no threshold below which the tool declines to answer, no statement that it says so when the text is too thin or falls outside its training, and nothing on what a user sees for an invention in a technology area the model has little data for. A 10,000-word input cap is documented, with excess text silently ignored.
Fabricated Citation Record
Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?
No court order, opinion or disciplinary record addressing fabricated or hallucinated legal citations produced by this product has been located as of the date shown. This is a statement about the public record on that one subject, not a finding about the product, and this signal is not a litigation history.
No record was located of this product's output being found fabricated or inaccurate in a proceeding, a regulatory action or a published account. Searches on 20 September 2026 across the vendor's estate, press and directory profiles returned nothing of the kind. The shape of the risk here is not a fabricated citation: the tool produces predicted art units, CPC codes and probability-style analytics rather than legal text or authority, so the failure that would matter is a prediction that steers claim drafting toward the wrong examination path. Nothing published describes such a case and no account of one was found.
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.
Nothing on the estate refers to lawyers' professional or ethical obligations. No rule of professional conduct, ethics opinion or bar guidance is named, and professional responsibility is not engaged in general terms either. The published Terms of Use carry no statement that the product gives no legal advice and create no position on an attorney-client relationship; what they carry instead is a disclaimer of accuracy and reliance.
This matters more than the usual absence because of who the product is sold to: patent practitioners are a named segment, and so are individual inventors, and the tool's purpose is to influence how claims are drafted before filing. The nearest material is a downloadable guide to hiring a patent attorney, which is practical advice about choosing counsel rather than anything about a practitioner's own duties. Checked the Terms of Use, the FAQ, the Predictor page, the About page and the full footer on 20 September 2026.
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, and the product sits inside a fee relationship between a lawyer and a client where those savings would change the bill.
The whole commercial argument is about legal fees, and none of it is about disclosure. The product is sold to the paying side of the relationship: a named use case is reducing outside legal services fees, the platform forecasts prosecution cost, flags a case running over budget or taking too many rounds, and benchmarks the firm's own counsel against examiner data so a client can ask why a case is off track. The vendor also positions its own price against enterprise IP platforms and describes an annuity feature it is building as far cheaper than routing payments through a law firm and an annuity service, though that feature is not yet available.
What is entirely absent is the other direction: nothing addresses whether a practitioner using the Predictor should tell a client that machine prediction shaped the claim strategy, or how work assisted by it should be billed.
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.
A buyer running a diligence checklist would come away with very little. No subprocessor list is published, no security certification or penetration test reference exists, no data processing addendum is offered, and nothing is available on request. The two documents that would normally answer these questions, the product privacy policy and the end user licence agreement, are named in the published Terms of Use but sit behind the application login and could not be read.
What is readable and useful is narrow but real: the Predictor discards submitted invention text once a prediction is generated, hosting is on United States based servers, access is role-based with portfolio-level isolation, and the models are the vendor's own trained on public USPTO data rather than a third party's. Checked the Terms of Use, the FAQ, the Predictor page and the full footer on 20 September 2026.
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.
Nothing published addresses disclosing the model's involvement to anyone outside the company. No court sits in this product's path, but the equivalent forum does: applications shaped with the Predictor are filed at the USPTO, and whether machine assistance in drafting or claim strategy needs to be disclosed to the examining office is a live question the estate does not raise. Inside the platform a record of sorts exists, since predictions can be rerun and the case history is retained and cannot be permanently deleted, but nothing describes exporting that trail, nothing distinguishes a claim amendment made on a prediction from any other edit, and no format or guidance is offered for any tribunal or agency. Checked the Predictor page, the FAQ, the help centre and the Terms of Use on 20 September 2026.