What did Anthropic announce on 28 October 2026?
Anthropic added a gallery of ready-to-use artifacts to Claude for Legal on 28 October 2026. The company calls them interactive tools, such as trackers and dashboards. Legal teams can adapt them with their own data and share them across a business. Legaltech News, the ALM trade publication, described the artifacts as reusable digital products. They sit alongside the chat window while a user works with Claude.
The release updates Claude for Legal, which Anthropic introduced on 12 May 2026 with a bundle of connectors and practice-area plugins. Anthropic said the most common requests from legal teams since May were "simpler setup and easier ways to work with colleagues".
Six artifacts form the core of the gallery. Each one opens with a worked example, so a lawyer can see how the output is assembled. Anthropic says every answer or output carries a link to its source, a point it repeats throughout the announcement.
Anthropic did not announce a separate price, and the product is listed under Claude Enterprise. Teams are directed to Anthropic's sales organisation or implementation partners when planning a rollout. Reuters reported the launch on the same day. The Decoder summarised it as pre-built templates for tasks legal teams had already requested.
Anthropic built the gallery around the tasks its legal customers raised most often. It also opened the gallery to outside builders instead of keeping the catalogue in-house. Law firms, legal tech vendors and legal aid groups can submit artifacts. Anthropic says several partners have more in development, and the release followed two other industry launches. Anthropic published a Claude for Financial Advisors update on 27 October and an expansion of Claude for Nonprofits on 26 October.
Outside builders matter because the catalogue is small. Six core artifacts cannot cover every legal task. Anthropic said in May that no single set of plugins covers every practice. The partner entries are partly a response to that gap. They also give outside vendors a route into a customer's Claude workspace.
What are the six artifacts in the legal gallery?
The six tools cover work that recurs across in-house departments and law firms. Most take a pile of documents or a stream of requests and turn them into a structured view a lawyer can review.
The division of labour matters for how these tools are sold, and Anthropic's description is simple. A legal professional brings their own documents and sources, and Claude completes the first pass. A lawyer then applies judgment to the result, so the artifacts are drafting and triage aids rather than decision-makers. Anthropic also says artifacts work best when connected to existing systems. The gallery is meant to be paired with connectors rather than used in isolation.
Anthropic's framing also puts the verification burden on the person using the tool. Source links are meant to make checking faster, but they do not remove the need to check. That distinction is central to how legal teams are likely to evaluate the gallery. Anthropic presents each artifact as a starting point that teams adapt with their own data and connectors.
- •Cite check compares a document with its sources, flagging quotes that do not match, wrong page numbers and unsupported claims.
- •Jurisdiction survey matrix compares a legal topic across states or countries in one table, with links to the underlying law.
- •Matter board turns engagement letters, emails or spreadsheets into one board of open matters, with owners, next steps and deadlines.
- •Playbook deviation board compares contracts with standard terms, flags each differing clause and routes exceptions to the approver.
- •Regulatory horizon scan maps upcoming regulations on a timeline and ranks them by impact, with links to the sources.
- •Tabular review builds a table from contracts, showing terms such as severance or non-competes, with links to the exact language.
Which legal tools and firms are connected to Claude for Legal?
Anthropic says it has added more than 60 legal connectors since the May launch. Named examples include Bloomberg Law, Filevine, Ivo, Legora and Patlytics. Connectors pull a firm's precedents, research sources and matter documents into Claude. Users do not have to upload files by hand.
The May launch set the pattern with more than 20 MCP connectors and 12 practice-area plugins. These cover commercial, corporate, employment, privacy and litigation work, among other roles. Claude also runs inside Microsoft Word, Outlook, Excel and PowerPoint. A redline drafted in Word can carry its context into an email or a closing checklist.
Four partner-built artifacts are in the new gallery. Goodwin contributed an AI Governance Benchmark, Harvey built Harvey Intake, and PwC built a Subpoena Response Tool. Foley & Lardner built an Obligations Checker, and Anthropic says Legora and Orrick have more in development.
Sharing is governed by settings that Anthropic says maintain privacy and support ethical walls between teams. The Decoder noted that Anthropic makes this claim rather than offering independent verification. Rollout partners include Accenture, Deloitte, KPMG, PwC, AnswerRocket, Caylent, Perficient and Sia Partners.
Several connectors target the document-heavy parts of legal work. Docusign connects agreement data so teams can surface renewal dates and obligations. Ironclad gives Claude access to a contract repository and scopes results to each user's permissions. Thomson Reuters connects CoCounsel Legal, which is grounded in Westlaw primary law and Practical Law guidance.
Permissions recur in Anthropic's connector descriptions, from Ironclad's per-user scoping to Box's existing security policies. Anthropic presents that design as a way to avoid bulk exports from a firm's document systems. For managing partners, the appeal is that Claude inherits access rules rather than bypassing them.
Why is Anthropic pushing into legal work now?
Anthropic has a simple commercial reason to keep investing in legal. At the May launch, it said legal professionals had become the most engaged Claude Cowork users of any knowledge-work function. Legal teams were already heavy users before dedicated tools existed.
The partner quotes in the announcement point to volume. Scott Stevenson, co-founder and chief executive of Spellbook, said the company is "approaching 500,000 contracts per month". Docusign chief executive Allan Thygesen said the integration lets legal teams and other staff create, review and move agreements forward. Mindy Lok, Accenture's legal chief technology officer, said Claude is "helping us transform the way our legal teams work".
Legal is also one of the few professional markets with large AI-native vendors. Legora reached a $5.6 billion post-money valuation in April 2026. TechCrunch reported that the company crossed $100 million in annual recurring revenue. Legora said its platform was used by more than 1,000 law firms and in-house teams across 50 markets. Reuters reported in September 2026 that Harvey reached a $15.5 billion valuation in a new funding round.
Harvey and Legora are partners as well as competitors, since both appear in Anthropic's connector directory. Anthropic can therefore sell into legal departments through software they already use. The risk cuts the other way, because model providers can move up the stack, and a gallery of artifacts is that move.
Anthropic also leans on partner benchmarks to argue that its models are strong at legal work. In May it cited Harvey's BigLaw Bench, where Claude Opus 4.7 scored 90.9%, "the highest of any Claude model" at the time. In October, Solve Intelligence's chief research officer reported a newer result. Claude Opus 5.5 "correctly identifies over 10% more structures than Opus 5" on the company's chemical benchmark. He added that it uses less than half the output tokens. Legal suits this pattern because the work is document-heavy and many vendors already built on Claude.
What do the access-to-justice artifacts do?
Nine of the gallery's artifacts are aimed at access to justice, and five were built by nonprofit and public interest organisations. Anthropic supported those groups with Claude credits, and several of their tools are tied to a single jurisdiction.
Zealous built a Criminal Charge Reality Check for defence teams. It shows the sentencing range for a charge and public data on how similar cases usually end. It also offers standard playbooks for release, and Anthropic states that it covers New York only. Mobile Pathways built an Immigration Papers Reader. It turns a client's paperwork into a plain-language first read of key dates and missing documents.
Rising for Justice contributed two tools based on District of Columbia law. A Rent Ledger Reviewer helps attorneys find errors in a landlord's ledger before an eviction case. A Record-Sealing Screener assesses whether a client's criminal record is eligible for sealing. The Access Project built a Record Clearance artifact. It helps legal aid teams work out whether a record can be cleared under state rules.
Anthropic tied the launch to its Oxford Justice Innovators Programme. That is a 12-month effort with the Oxford Institute of Technology and Justice and Microsoft. It supports up to 10 innovators, and Anthropic also opened a waitlist for a Legal Ambassadors Program. The programme is a community for legal professionals who build with Claude. The jurisdiction limits matter because a tool built for New York cannot simply be pointed at another state's statutes. Scott Hechinger of Zealous said the tool "started with one question". The aim was to speed up the arithmetic so lawyers could spend time on the person instead.
What are the risks and limitations for lawyers?
The central limitation is accuracy, and the evidence base is sobering. A Stanford RegLab study released in 2024 tested retrieval-augmented legal research tools. It covered products from LexisNexis and Thomson Reuters. The authors found that the tools hallucinated between 17% and 33% of the time. They described providers' claims of hallucination-free citations as overstated.
Court records show the practical cost. The AI Hallucination Cases Database is maintained by Damien Charlotin. It logged 2,241 judicial decisions worldwide as of 29 October 2026. Of those, 1,514 were in the United States, 238 in Canada and 89 in the United Kingdom. Some 1,839 entries involved fabricated material. The database counts decisions where a court found or implied that a party relied on hallucinated content. It is a floor rather than a full measure of the problem.
Professional rules add a second layer. The American Bar Association issued Formal Opinion 512 in July 2024. It was the association's first ethics guidance on generative AI. The opinion tells lawyers to weigh competence, confidentiality, client communication, fees and supervision. A link back to a source supports that duty but does not discharge it. The case database also tracks sanctions and disciplinary referrals.
There are structural questions too. Artifacts are shared across a business. That means a legal team's work product may sit in front of colleagues who are not lawyers. Anthropic says sharing settings enforce privacy and ethical walls. In law firms, information barriers are usually legal and contractual obligations rather than product settings. Anthropic has not published accuracy figures for the artifacts themselves. The announcement also does not describe how a firm would audit their outputs.
Cost and procurement are unresolved in public. The announcement gives no per-seat price and no usage limits. It does not say which Claude Enterprise tiers include the gallery. Firms will need those answers before comparing the tools with dedicated legal platforms, which already publish benchmarks of their own.
Benchmarks in this market are usually run by the vendors selling the tool. Harvey operates BigLaw Bench, and Solve Intelligence runs its chemical-structure test. Both companies build on Claude. Those figures are evidence of partner confidence. They are not a substitute for independent evaluation of the artifacts that shipped this week.
One practical gap is supervision. A partner reviewing an artifact's output sees a table or a board, not the reasoning that produced it. Anthropic says every output links to its source, which helps, but the volume of links in a tabular review can be large. Firms will need their own review routines for anything that leaves the building.
What should legal teams watch next?
The first thing to watch is the partner pipeline. Anthropic says Legora and Orrick have artifacts in development. The gallery is designed to accept submissions from legal tech companies, law firms and legal aid groups. If outside organisations can publish trusted artifacts, the gallery becomes a distribution channel.
The second is measurement. Anthropic has not released accuracy data for the six core artifacts, so independent testing is the missing piece. Watch whether law schools or competitors publish evaluations of cite check and tabular review. Watch whether courts start citing artifact output in sanctions decisions.
The third is competitive response. Harvey, Legora and Thomson Reuters already sell legal workflows built on Anthropic models. Each has an incentive to keep its own interface in front of the lawyer. Anthropic's gallery puts a basic version of that interface inside Claude. The boundary between model provider and application vendor becomes harder to maintain.
Finally, watch pricing and rollout partners. How partners package Claude for Legal will shape who can buy it. Integration fees will decide whether the artifacts reach small in-house teams. Two open questions sit underneath all of this. Will Anthropic publish accuracy data for each artifact? Will the legal aid tools expand beyond their first jurisdictions?