Huski.ai

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 image recognition model is the product and the founding insight, not a layer over an existing service. The company was built in 2020 around applying visual similarity matching to trademark work, and the vendor states the technology eliminates the need for design search codes, which is the precise claim: the USPTO design coding system exists because computers could not compare images, and a model that compares them directly replaces the taxonomy rather than accelerating a search within it. The same recognition engine runs across all four products, matching filed applications in watch, clearing candidate designs in search, and identifying infringing product listings in e-commerce including images the vendor states are blurred or obscured. Remove the model and there is no search, no watch and no brand protection, only a docketing tool. Distinguished from Corsearch at B in this category: Corsearch is a decades old search and services business that added AI, and this vendor exists because of it.

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 mechanism is described and demonstrated, and nothing measured is published. Independent coverage from a named legal technology journalist describes running an image search against the USPTO database and receiving visually similar results within seconds, and reproduces a worked example in which the tool identified a filed design as similar to an existing Louis Vuitton mark. A demonstrated example in third party coverage is better evidence than a claim, and it is credited as such rather than as a measurement. What is absent is any figure: no precision or recall on visual similarity matching, no false negative rate on clearance, no comparison against design search code retrieval which is the incumbent method the vendor says it replaces, no evaluation and no test set. The gap is consequential for the same reason recorded on Corsearch: in clearance the failure mode is a silent miss, and a confusingly similar mark the model does not return produces an opinion that is wrong and looks complete. Checked the home page, the solutions pages, the FAQ and independent coverage on 29 Aug 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 professional is structurally in the loop and no oversight model is published. The product is sold exclusively to trademark attorneys, paralegals and legal professionals, with the vendor stating plainly that IP attorneys and their staff are the end users, so every output reaches a qualified reviewer by design rather than by policy. Output is framed as input to professional work throughout, with search results feeding a report and a customised opinion letter the attorney issues, and watch results feeding a decision on whether to file a Letter of Protest or opposition. What is not published: no statement of what runs unattended, no confidence indication on a visual similarity match, no threshold governing what surfaces in a watch alert, and no description of what happens when the recognition engine is uncertain about a blurred or obscured listing image in brand protection, where an automated conclusion could support an enforcement action against a third party seller. Structural oversight is real and is not a substitute for a published model.

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.

Independent journalism and a demonstrated product, with no customer evidence at all. The strongest element is coverage by a named legal technology publication that reviewed the product at launch, described the founders and their reasoning, ran the search itself and published a worked example of a result, which is independent verification that the core capability functions as described. Company facts are checkable: founded 2020, based in Middletown, Delaware, with named cofounders. Independent directory and review listings exist. Against that: no law firm or customer is named anywhere in located material, no case study, no usage figure, no funding announcement and no award were found, and no outcome or efficiency claim is published with a figure beyond the pricing comparison. Held at C rather than D because independent hands on coverage is a materially better evidence base than the self authored material that left IP Author at D in this category, and below B because nothing establishes adoption.

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.

Nothing located. The product holds candidate marks and unlaunched brand designs uploaded for clearance, which are commercially sensitive before filing, and it generates opinion letters, which are the archetypal attorney work product in trademark practice and the document that becomes contested in later infringement litigation on the question of what the client knew before adoption. Nothing published addresses attorney client privilege, work product, the confidentiality of a clearance search or of a generated opinion letter, or the conflicts question that arises when one platform serves many firms clearing marks against each other's clients. The vendor's exclusive focus on legal professionals makes the omission more surprising than it would be for a general enterprise tool. Checked the home page, the solutions pages, the FAQ and the site navigation on 29 Aug 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.

A clear scope statement exists and no professional framework sits behind it. The vendor states unambiguously that Huski is for trademark attorneys, paralegals and other legal professionals, that IP attorneys and their staff are the end users, and that clients benefit indirectly, which is an explicit restriction of the user population to qualified professionals and the reason this grades above the D recorded on most of this index. Selling only to lawyers is a real answer to the unauthorised practice question, since the tool never reaches the public. What is absent is everything else: no statement that generated output including the customised opinion letter is not legal advice, no positioning on the attorney's duty of competence over a machine generated opinion letter issued under their own name, and no engagement with any bar guidance. Generating an opinion letter in a few clicks is the sharpest version of that question in this category, and it is not addressed.

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.

Nothing published about how the models are governed, evaluated or monitored. No AI policy, no model card, no bias or fairness testing, no evaluation methodology, no accuracy monitoring, no drift statement, no named governance body, no ISO 42001 and no EU AI Act positioning were located. The untested risk is specific to image recognition and worth naming rather than listing generically: visual similarity models are known to perform unevenly across image types, and this engine is applied to logos, product photographs and deliberately obscured listing images alike, with results feeding both clearance opinions and enforcement actions against third party sellers who never see the model. Nothing published indicates that performance across those conditions has been examined. Checked the home page, the solutions pages, the FAQ, the published rankings material and the site navigation on 29 Aug 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.

No stewardship position located. Nothing states whether uploaded candidate marks, logo images, client portfolios or search histories are used to train or improve the recognition models, no retention period is published, and no deletion right is described. The gap has a distinctive edge on this product: the platform ingests unfiled logo designs at the clearance stage, which are both commercially sensitive and precisely the kind of labelled image data that improves a visual similarity model, so the incentive alignment is visible and the position is unstated. Docketing additionally holds entire client portfolios uploaded in one click, and the Customs and Border Protection integration means recordation status data flows through the platform. Checked the home page, the solutions pages including docketing and brand protection, the FAQ and the site navigation on 29 Aug 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 published position located. Nothing was found on liability for AI output, warranty, service levels or remedy. Three distinct exposures arise from this product and none is addressed. A clearance search that misses a confusingly similar mark supports an opinion letter the attorney signs, and the opinion letter is the artifact a client relies on to launch a brand. A docketing system that fails to surface a deadline causes a lapsed registration, which is a classic malpractice claim and is not covered by any published service level. And a brand protection match that wrongly identifies a legitimate listing as counterfeit supports an enforcement action against a third party seller with no described recourse. Checked the home page, the solutions pages, the FAQ, the pricing material and the site navigation on 29 Aug 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.

One genuinely uncommon integration and none of the ordinary ones. The vendor states integration with US Customs and Border Protection to track recordation statuses, which connects the platform to the government system through which recorded marks are enforced at the border, and no other record on this index names a customs or enforcement authority integration. One click portfolio upload with automatic population of clients and trademarks from filing record details implies a working connection to USPTO filing data rather than manual entry. What is absent is everything a firm would ask about: no IP management or docketing system integration, which matters less here because this product is itself the docketing system, no document management system, no practice management connector, and no API or export documentation was located beyond single click report generation. Checked the docketing page, the solutions pages, the home page and the site navigation on 29 Aug 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 located. No hosting provider is named, no region or data residency commitment is published, and no deployment options are described. The vendor is a small United States company serving United States trademark practice against the USPTO, so residency is a narrower question than for the multinational vendors elsewhere in this category, and it remains unanswered for a platform holding unfiled logo designs and complete client portfolios. Checked the home page, the solutions pages, the FAQ and the site navigation on 29 Aug 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, attestation, trust centre or security page was located. No SOC 2 of either type, no ISO 27001, no named auditor, no penetration testing partner, no encryption statement and no security documentation of any kind was found across the pages read. Under the three tier test the artifact is absent rather than gated. Two records in ip-and-patents now sit at D on this axis, this one and Corsearch, and both are the trademark vendors, while all five patent vendors publish at least one named certification. Recorded as an observation rather than a conclusion, since two records is a small sample, and flagged for the category note. Held at D rather than higher because unlike IP Author, where a published Trust Center existed but could not be read, no trust surface of any kind was located here to hold a grade open for. Checked the home page, the solutions pages, the FAQ, the pricing page and the site navigation on 29 Aug 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 located. No foundation model provider, model family or version is named, no distinction is drawn between proprietary and third party models, and no subprocessor list was found. The vendor describes advanced image recognition and AI powered brand recognition throughout, which characterises capability rather than provenance, and given the image recognition engine is the entire technical proposition, whether it was built in house or licensed is the single most material supply chain question about this product and it is unanswered. Checked the home page, the solutions pages, the FAQ and independent coverage on 29 Aug 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.

THE SECOND A ON THIS AXIS IN 54 RECORDS, after Descrybe, and the only one in ip-and-patents. Everything a buyer needs is published without contacting anyone: a stated entry price of $299 per month, a dedicated pricing page, the unit of charge stated as unlimited text and image clearance searches rather than metered per search, discounts stated for twelve month commitments, and explicit terms of no contracts and no credit card required to start. The vendor also positions its price against the market openly, and independent coverage corroborates the comparison, noting that competing trademark search tools commonly charge $100 to $800 per search. A firm can calculate total cost of ownership and compare it against incumbents from published material alone. Graded A because that is the top of this axis. Held short of a perfect record because the full tier structure above the entry price was not read in this pass and it is unclear which of search, docketing, watch and brand protection the entry price includes.

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.

Lifecycle coverage is complete and jurisdictional coverage is narrow and largely unstated. Within its scope the product spans the whole trademark lifecycle in one platform: clearance search, docketing with deadline management, watch on new filings, Office Action handling, and post registration enforcement through e-commerce brand protection, which is broader than most single vendors in this category attempt. E-commerce coverage is characterised specifically, naming Shopify powered independent stores as a claimed unique capability with the reasoning stated, that Shopify has no centralised marketplace and over three million stores operate on custom domains. Held at C because jurisdictional coverage appears to be United States only: the USPTO is the sole trademark office named across search, watch and docketing, the Customs and Border Protection integration is United States specific, and no international registry, Madrid Protocol handling or foreign filing coverage was located. Compare Corsearch at A in this category with 190 registries and 1,100 databases stated. A single office platform is a legitimate design choice and is a real coverage limit for any firm with international clients.