LegalOn
Contract review and legal workflow platform for in house legal, legal operations, procurement and contracting teams, built around attorney written playbooks rather than customer built ones. Ships more than 135 pre built playbooks covering around 10,000 legal issues across contract types, negotiating positions and dozens of countries, maintained as law changes, with a Playbook Agent that converts a customer's own templates and past redlines into structured playbooks. Other surfaces are Assistant, Review and redline in Word or browser, Matter Management with an intake and triage agent, Vault for signed contract storage and extraction, Translation across 28 plus languages, and entity and board management. Publishes an annual contract review benchmark measuring its own system against foundation models. Operated by LegalOn Technologies, which originated in Japan and now runs a US and global business alongside its Japanese one.
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 product. Review, redlining, Assistant, the intake and triage agent, the Playbook Agent, translation and Vault extraction are all generative or machine learning capabilities. The vendor's own framing is that it builds a harness on top of foundation models: legal content, review architecture, evaluation design and model orchestration. Remove the models and what remains is a library of attorney written playbooks with nothing to execute them.
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 most complete measurement disclosure on the index. The vendor publishes a 2026 Contract Review Benchmark testing 11 named AI models against its own system across 3,282 pairwise reviews on 21 precision critical guidelines, with the methodology described: contracts are broken into structured provision level checks, each evaluated against a precise legal standard, scored by an LLM judge whose outputs were independently validated by legal experts against professional standards. Failure modes are named rather than implied, the finding being that general purpose models from Anthropic, Google and OpenAI produce confident sounding answers that are frequently wrong on provision level review. Speed is measured alongside accuracy. The report is downloadable and the vendor states its own scepticism standard for vendor benchmarks in publishing it. Output grounds to the playbook standard with citations and clickable references to source. The obvious caveat, that this is the vendor's own benchmark of its own product, is disclosed by the vendor itself and does not pull it below the A band, which asks for named, dated, checkable evidence rather than third party evidence.
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.
A real published commitment with described review surfaces, short of the full control structure. The vendor states the design principle explicitly: it is built to hand off what a lawyer would hand off and hold the line where a lawyer would hold it, every output is sourced and easy to verify, and final judgment and control stays with the user. The intake agent has a described escalation rule: standard paper moves on its own and only real exceptions reach a lawyer. Review returns pass or fail indicators with risk ranked low, medium or high and clickable references, and redlines land as tracked changes a lawyer accepts or rejects. Not located as of 29 Aug 2026: where the threshold sits that separates standard paper from an exception, who configures it, and what the vendor commits to when an agent is wrong.
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.
Named customers with figures and a method a reader can assess. Multiple published case studies pair a named organisation with a quantified result and the work it applies to: TriHealth at 75 percent reduction in contract review time on clinical trial agreements and vendor contracts, Industrial Service Solutions saving a lead reviewer 15 plus hours a week across named agreement types, a named software customer at 70 percent reduction in NDA review time, and MCEA Holdings at 50 percent more contracts handled without added headcount with senior oversight now under 10 percent. Individually attributed quotes come from named counsel at named organisations including TriHealth, Onshore Companies, Astrophysics and Tekscend Photomasks. The logo wall names large enterprises including Panasonic, Toshiba, Suntory, Ingersoll Rand, Wikimedia and the University of Washington, and the vendor states more than 9,000 legal teams. Short of nothing material: the figures are customer reported rather than independently audited, which is normal for this evidence class.
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.
Substantive published commitments, short of the full picture. The training position is specific and covers the model provider layer: the vendor works with Microsoft's Azure OpenAI service under protections that no customer data is used to train, retrain or improve those models and that no customer data is stored for any period, explicitly including for abuse monitoring and debugging. That last clause is a real detail most vendors omit. Segregation is documented as segregated environments per customer account with strict access controls, which meets the applicable level for an in house buyer under the amended band. Encryption at rest and in transit is stated. Two gaps hold this off an A. Attorney client privilege and work product handling is not addressed directly in located material. Retention and deletion of customer documents in the vendor's own systems, as distinct from the model provider layer, is not stated.
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 boilerplate disclaimer sits in the terms while the product is sold to non lawyers by design. A disclaimer is published in the terms and conditions, and the vendor's framing of holding the line where a lawyer would hold it is a real design statement. But dedicated solution pages target procurement, sales, contract management and legal operations, and vendor material states directly that AI contract review is not only for lawyers and that procurement managers, sales contract teams and compliance officers use it without routing everything through legal. Searched the site, the solution pages, the published terms and conditions and the resource hub on 29 Aug 2026 and located no position on the advice line, no treatment of competence or supervision duties, and no statement of jurisdiction limits, despite playbooks covering dozens of countries.
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.
A published governance framework with real substance, short of a named owner or bias testing. What is published goes beyond principles: a described evaluation regime with an LLM judge validated against attorney assessment, a named internal AI function through published contributors including a VP of AI and a data scientist, an attorney authored content pipeline with playbooks maintained as law changes, and a stated commitment to ongoing transparent evaluation across future task areas including redlines, research, extraction and drafting. Testing is therefore both described and executed rather than asserted. Not located as of 29 Aug 2026: pre release testing gates as distinct from published benchmark results, a named accountable owner for model governance, and any disclosure about uneven output across matter types, parties or populations.
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.
Substantive published policy covering most of the ground. Published: segregated environments per customer account with strict access controls, encryption at rest and in transit over public networks, hosting named as AWS and Google Cloud Platform in the US, annual penetration testing plus regular vulnerability testing, a described patch management process, and zero storage at the model provider layer including for abuse monitoring. A trust center is published at a stable URL. Not located as of 29 Aug 2026: a stated retention period or deletion control for customer documents in the vendor's own systems, a named subprocessor list, and an incident or breach notification practice. Retention is the notable absence given Vault is designed to hold every signed agreement indefinitely.
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 standard limitation and disclaimer structure. Terms and conditions are published openly and carry a dedicated disclaimer section, so a buyer can read the allocation of loss before entering a sales process, which keeps this above a pure absence. Searched those terms, the privacy policy, the security page and the trust center entry point on 29 Aug 2026 and located no indemnity running to the customer for third party claims arising from output, no warranty on output, no stated liability cap figure and no insurance position.
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 integrations exist and are documented, short of implementer level depth. A Microsoft Word integration has its own product page and is where review and redlining happen, browser review is supported for docx and PDF, and the intake agent ingests requests from email, Slack or a web form and routes them. Matter management, Vault storage and entity management mean the product covers more of the workflow natively rather than integrating to it. Not located as of 29 Aug 2026: legal document management connectors such as iManage or NetDocuments, contract lifecycle or e signature connectors, and per integration documentation of what moves in which direction and what an administrator configures. No integrations index page was located.
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.
Cloud delivery is stated and the region is fixed rather than offered. The vendor publishes that hosting runs on AWS and Google Cloud Platform in the US, which is a real residency statement, and states segregated environments per customer account, which speaks to isolation. What is missing is choice and detail: no selectable region, no single tenant or private deployment option, and no statement of where processing happens as distinct from where data is stored. For a vendor with a Japanese parent business, playbooks across dozens of countries and stated GDPR compliance, the absence of any EU or Japan processing option on the global site is a live buyer question the page does not address.
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.
Certification is real and stated with an open route, short of accessible evidence. SOC 2 Type II is stated as certified with the standard correctly attributed to the AICPA, and ISO/IEC 27001:2022 and ISO/IEC 27017:2015 are both named with their versions and with what each covers explained, the 27017 description correctly identifying it as cloud specific controls for multi tenant and virtualised environments. Annual penetration testing is stated. A trust center is published at a stable URL, a self serve route rather than a sales gate. Two things hold this off an A. No coverage period, audit scope, report date or named auditor was located as of 29 Aug 2026. And the home page certification strip lists ISO 27007, which is the guidelines standard for auditing information security management systems and is not a certifiable standard for an operator, while the security page itself correctly says 27017. Recorded as a discrepancy between two pages of the same site rather than as a conclusion about the vendor.
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 supply chain is partly disclosed and the disclosure is specific where it exists. The vendor names Microsoft's Azure OpenAI service as a model provider and states the contractual protections attached to it, and its published benchmark names Anthropic, Google and OpenAI models as the systems it tests against and states that its own harness is built on top of foundation models, with Claude Opus 4.6 named in coverage of that benchmark. Hosting is named as AWS and Google Cloud Platform. What is missing is the direct question: which model serves the product's own live output, since the named models appear as benchmark comparators rather than as a stated production stack. Not located as of 29 Aug 2026: a subprocessor list, and any commitment to notify customers when the model supply chain changes.
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.
Checked the pricing page, the home page, the platform and solution pages and the footer on 29 Aug 2026. A pricing page exists in the navigation but publishes no rate, no unit of charge, no tier structure and no seat minimum. Every commercial path on the property terminates in a demo request or a sales phone number, which is sales gated and earns no credit. No free trial or self serve entry point was located. Third party pricing figures for this vendor were not located either.
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.
Who the product serves is documented precisely, with the boundaries visible in the structure rather than only claimed. Six buyer functions carry dedicated pages: in house legal, legal operations, procurement, contract management, sales and law firms. Six verticals carry their own: construction, financial services, business services, private markets, healthcare and manufacturing, with software named separately. Practice coverage is stated at the level of contract type and negotiating position rather than in general terms, with more than 135 playbooks across roughly 10,000 legal issues, each specific to whether the customer is sending or receiving the paper, and international playbooks giving jurisdiction specific guidance across dozens of countries. Translation covers 28 plus languages. The self limit is clear from the structure: this is contract and contracting work, not litigation, and nothing on the property claims otherwise.
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?
A public policy or trust page states no training on customer content, with no matching term located in the published agreement.
The security page states the position at the model provider layer with unusual specificity: working with Microsoft's Azure OpenAI service under protections that no customer data is used to train, retrain or improve those models, and that no customer data is stored for any period of time, explicitly including for abuse monitoring and debugging. The home page adds that nothing shared with the vendor ever trains a third party model. Two limits on what that covers, both recorded rather than resolved: the commitment as written addresses third party models rather than the vendor's own systems, and it was located on the security page rather than in the published terms and conditions, which were searched on 29 Aug 2026.
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.
At the model provider layer the answer is zero and is stated plainly, including a specific carve out that storage does not occur even for abuse monitoring or debugging, which closes a gap most vendors leave open. What was not located as of 29 Aug 2026, after checking the security page, the published terms and conditions, the privacy policy and the trust center entry point, is any retention period for the vendor's own storage, any customer control over that window, or any deletion commitment. That gap matters more here than for most: Vault is a product designed to hold every signed agreement as a permanent searchable record, so a buyer is being asked to store contracts indefinitely with no published retention or deletion terms.
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.
The product maintains its own documented segregation model rather than inheriting one from a document management system. Segregated environments per customer account with strict access controls are stated, and ISO 27017 certification is described as covering cloud specific controls for multi tenant and virtualised environments, which speaks to the same boundary. That is tenant level separation, documented. What is not addressed is segregation between users or matters inside a customer, and no legal document management integration was located whose permissions retrieval could enforce at query time. The product is sold primarily to in house teams, where tenant level separation is the applicable question, but a dedicated law firm solution page also exists.
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.
Searched the published terms and conditions, the privacy policy, the security page and the trust center entry point on 29 Aug 2026. No clause committing to notify a customer of a government or law enforcement request for their data was located, and no transparency report was located. This records a search across the published documents that did not surface the clause rather than a reading of every document end to end.
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 here is attorney written playbook content rather than primary law, which is the right shape for a contract review product and reads differently against a signal written for case law. Provenance is unusually well stated for what it is: the vendor names its own attorneys as the authors, states the content is built and maintained in house and kept current as laws and standards change, and quantifies it at more than 135 playbooks across roughly 10,000 legal issues covering dozens of countries. Because the vendor authored the content, the licensing question that this signal exists to probe largely does not arise. What is not published is the update cadence in concrete terms, or which underlying legal sources the attorneys drew on when writing the standards.
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.
Searched the site, the Review and Playbooks product pages, the knowledge core page and the resource hub on 29 Aug 2026. No material was located addressing whether authority carries a treatment signal or whether subsequent history is checked. The vendor does state that playbooks are maintained as laws change, which is the nearest equivalent for this product type and is a currency commitment about its own content rather than a citator. Noted for context: this is a contract review product whose grounding is playbook standards rather than case law, so a citator is largely outside its design.
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.
The product exposes a structured confidence signal rather than an abstention path. Review returns pass or fail indicators per playbook item with risk ranked low, medium or high, and clickable references to the source or to suggested locations, so a reader sees how strongly a provision is flagged and can check it. Searched the site, the product pages, the benchmark material and the resource hub on 29 Aug 2026 and located no explicit no answer path, and no published statement of what the product does when it cannot ground an assessment.
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.
No court order, opinion or disciplinary record naming this product has been located as of 29 Aug 2026. Instrument searched: the AI Hallucination Cases database maintained by Damien Charlotin, which tracks court decisions worldwide involving AI generated hallucinated content and records the AI tool implicated where it is known. Also checked published 2026 sanctions summaries and secondary sanctions trackers. The entries located name filers, and in some rows other products, rather than this one. This is a statement about the public record on the date shown and not a clearance. Note that the product is a contract review tool rather than a litigation or research tool, so its output is unlikely to reach a court filing as cited authority in the first place.
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.
Searched the site, the blog, the resource hub, the newsroom and the published guides on 29 Aug 2026. No engagement with any named ethics opinion was located, including ABA Formal Opinion 512, any state bar guidance, or any Japanese bar association guidance given the parent company's origin. The vendor publishes substantial research and survey material about how legal teams adopt AI, and an evaluation programme, both of which address performance and adoption rather than the professional responsibility obligations its buyers are bound by.
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.
Savings are claimed prominently and quantified, with nothing published on the client side of the equation. Published figures include 85 percent faster contract review, a 75 percent reduction in review time at a named healthcare customer, 15 plus hours saved per week for a named lead reviewer, and a customer quote about saving thousands in outside counsel costs. Searched the site, the customer stories, the resource hub and the published terms on 29 Aug 2026 and located no per matter record of AI assisted work intended for fee purposes, and no guidance on billing, fee or client disclosure treatment. Noted for context: the primary buyer is an in house team that does not bill a client by the hour, though a dedicated law firm solution page exists and that buyer does.
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 trust center is published at a stable URL and the security page names certifications, hosting providers and the model provider protections, which is real material a buyer can read without a sales conversation. But the specific artifacts this signal turns on were not located as of 29 Aug 2026: no subprocessor list, no consolidated statement of which model providers see customer content beyond the single named Azure OpenAI relationship, and no client facing consent or notification pack a firm could forward to its own client. Recorded as not addressed rather than at the subprocessor value because no list exists to point to.
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.
Searched the site, the product pages, the matter management and Vault pages and the resource hub on 29 Aug 2026. Review output carries citations and clickable references to source, and matter management records cycle times and workload, so elements of a trail exist. But no per document export covering model used, sources retrieved and human verification together was located, and the model used is not identifiable from published material in any case. Noted for context: this is a contracting product rather than a litigation product, so a judicial AI disclosure order is less likely to reach its output.