On 28 October 2026 Anthropic added a gallery of ready-made legal tools, which it calls artifacts, to Claude for Legal. Six were built by Anthropic, four by Harvey, Goodwin, Foley & Lardner and PwC, and nine are free tools for legal aid work. Each one opens with a worked example, and you swap in your own contracts, emails and connected apps. Anthropic's post about it on X passed 7.6 million views and 20,000 bookmarks within two days.
The tool lawyers care about most is cite check. Anthropic says it "compares a legal document with its sources and flags quotes that don't match, wrong page numbers, and sources that don't support the claim." It does not name any legal database behind it, and it publishes no accuracy rate. Read plainly, cite check checks your brief against the sources you load or connect. It is a check against your own sources, not a citation database like Westlaw or Lexis.
I think this gallery is the most practical thing Anthropic has built for lawyers. It gives a busy partner something that works on the first click. But the name "cite check" promises more than Anthropic's own words back up. And the promise that teams can work on these "with clients and partners" mostly stops at your firm's door.
What did Anthropic launch for lawyers?
Claude for Legal itself launched on 12 May 2026. It came with more than 20 connections to legal software and 12 plugins for different areas of law. This week's update puts ready-made tools on top. These are the six Anthropic built:
- Cite check compares a document with its sources and flags quotes and pages that don't match.
- Jurisdiction survey matrix compares one legal topic across states or countries in a single table.
- Matter board turns emails and spreadsheets into a list of open matters with owners and deadlines.
- Playbook deviation board flags contract clauses that differ from your standard terms.
- Regulatory horizon scan maps upcoming rules on a timeline and ranks them by impact.
- Tabular review pulls the terms you choose from a stack of contracts into one table.
The partner tools are an AI governance benchmark from Goodwin, a legal intake desk from Harvey, an obligations checker from Foley & Lardner and a subpoena response tool from PwC, with more coming from Legora and Orrick. The nine access-to-justice tools cover things like rent ledgers, record sealing and immigration papers. Five are named, and four of them are tied to US law. One is for New York only, two are based on Washington D.C. law, and one works under each US state's own rules. The fifth gives a plain-language first read of immigration papers. Anthropic supported the nonprofits that built them with Claude credits.
Anthropic also says it has added more than 60 legal connectors since May, including Bloomberg Law, Filevine, Ivo, Legora and Patlytics.

How much does Claude for Legal cost?
There is no separate price. Claude for Legal runs on normal Claude plans, and the "$20 a month" some sites quote is the price of Claude Pro. Anthropic's help centre says artifacts in general open on every plan, Free included, though Anthropic does not say which plans get the legal gallery. Connectors and saved storage, the two things that let an artifact work on your own files, need a Pro, Max, Team or Enterprise plan. The announcement is filed under Claude Enterprise, and Anthropic's legal page talks about Team and Enterprise customers, where it says it does not train on customer content by default.
What does Claude's cite check do?
Anthropic's description is careful. Cite check compares the document with its sources, each answer links to its source, and "a lawyer applies their judgment to the results." That is useful. If your brief quotes page 14 of a judgment and the quote is on page 41, or the words are slightly off, cite check is built to catch it.
What it cannot do on its own is tell you whether a case exists. If a fake case is cited and you never loaded it, there is nothing to compare it with, so the best cite check can do is flag the claim as unsupported. Tools that check against a real case law database exist inside Claude, but they are separate. The CourtListener connector from the Free Law Project checks citations against its database of US case law at no cost, Descrybe offers a similar check for $25 a month, and firms that pay for Westlaw can connect it too.
This matters because fake citations are now a flood. Damien Charlotin's database of court decisions involving AI-made-up material listed 2,241 cases worldwide on 29 October 2026. Lawyers were involved in 886 of them and people representing themselves in 1,292. Sixteen are from India.
One of them involved Claude. In Brooks v. Lowe's Home Centers, decided by a federal court in Louisiana on 18 May 2026, a lawyer used Claude to write a brief. A law clerk caught made-up quotations, so the lawyer asked Claude to fix them, then filed the new version without reading it. "Instead, Claude just made up more stuff," the judge wrote, and sanctioned him $1,000.

Can you share Claude legal artifacts outside your firm?
Only partly. Anthropic's announcement says artifacts "enable collaboration inside and outside the organization", and that a legal team can work on one "with clients and partners". Its own help centre, in pages dated 23 September 2026, adds the limits.
You can invite up to 50 people from outside your organisation to an artifact, and each of them needs a Claude account. For those outside people, "Parts of the artifact that use Claude or your connectors don't work for them." Artifacts that use Claude or connectors also cannot be shared with an "Anyone with the link" setting. So if you build a cite check or a tabular review and send it to a client or to co-counsel at another firm, the live parts arrive switched off. Outside invitations are on by default on Team plans and off on Enterprise plans, so a small firm on Team can share outside without anyone in IT ever deciding to.
Anthropic says its sharing settings "support ethical walls between teams", but it does not explain how. The closest the help centre comes is one line, that an artifact pulling from connected apps "uses the viewer's connections, not yours". It also gives its own warning: "Treat someone else's artifact the way you'd treat a file from an unknown sender."
There is one more line firms should read. Anthropic's help page on privilege and confidentiality, dated 26 August 2026, says that "for our most capable models, where safety monitoring requires it, we retain prompts and outputs for 30 days" in a store that is deleted automatically. It does not say whether zero data retention deals override that. A firm that assumes nothing is kept should get that answer in writing.
Claude for Legal vs Harvey: are legal AI startups finished?
No, and the money says so. Harvey raised $550 million at a $15.5 billion valuation in September 2026, and Legora raised $550 million at $8.8 billion on 21 October 2026. Both rounds came months after Claude for Legal launched. Harvey even built one of the artifacts in Claude's gallery, and Legora has one coming.
The big AI companies are still crowding in. Google launched Gemini Enterprise for Legal on 25 August 2026, OpenAI launched Astra for Law on 17 September, and Microsoft launched Copilot for Legal on 15 October. Anthropic's gallery came two weeks after Microsoft's. When Anthropic released plugins including a legal one at the end of January, Thomson Reuters shares fell as much as 18 percent on 3 February, on track for their worst day on record, and by May the company had become a connector partner instead. My read is that the giants are fighting over where a lawyer starts the day, while the specialists keep the deep workflows and the data deals.
What this means for India
Anthropic said in February 2026, when it opened its Bengaluru office, that India is Claude's second-largest market. But every named tool in the new gallery that is tied to one place is tied to the US. We also found no Indian legal database, such as SCC Online, Manupatra or Indian Kanoon, in the connector lists Anthropic has published. The jurisdiction matrix and horizon scan can be pointed at India, but no Indian case law connector from Anthropic sits behind them.
That gap matters more in India than anywhere else right now. The Supreme Court's draft rules on AI in courts, published on 3 June 2026, would make parties declare AI use when they file, let courts ask which tool was used and how the work was checked, and put full responsibility on the lawyer. The Bar Council of India has told Parliament that "the machine neither signs the pleading nor owes duties to the court." Big Indian firms are already using AI, with Cyril Amarchand Mangaldas on Legora and Shardul Amarchand Mangaldas on Harvey. An Indian lawyer running Claude's cite check today has to load every judgment by hand or bring their own connector, because Anthropic's lists have no Indian database to check against.
"Cite check is the fix for fake citations"
That is the strongest case for this launch, and it deserves a fair hearing. The mistake courts keep punishing is unchecked AI output, and cite check attacks exactly that by reading the real sources and flagging quotes and page numbers that don't match. If the Brooks lawyer had run his brief against the actual cases, the made-up quotations should have been flagged. Anthropic designed the gallery so a lawyer reviews every result, and Claude can connect to CourtListener or Westlaw for a database check. Courts, and India's Bar Council, already say the lawyer is responsible, not the software.
I agree with every part of that, and I still think the name is a problem. Cite check can only match what is in front of it, so the best it can say about a made-up case that was never loaded is that the claim is unsupported, not that the case does not exist. Anthropic has not published how often cite check misses a bad quote, so nobody outside Anthropic knows its error rate. And a tidy table of results looks finished in a way a chatbot reply does not. Brooks shows what happens when a lawyer trusts a confident machine and stops reading. Treat it as a check against your own sources, run a real database check as well, and read the sources yourself.
What Anthropic should do
Anthropic should test cite check on a public set of briefs and publish how often it misses a wrong quote or page number. It should also say plainly which sources the tool checks against. And it should put the sharing limits in the announcement itself. It tells lawyers they can work on artifacts "with clients and partners", but the live parts stop working for people outside the firm, and lawyers should hear that from Anthropic before a client tells them.
My prediction
By 30 June 2027, Claude for Legal will connect to at least one Indian legal database, such as SCC Online, Manupatra or Indian Kanoon. If I am wrong, I will say so here.
India is Claude's second-largest market. The Supreme Court's draft rules push lawyers to verify AI work before they file. And Anthropic already works with Adalat AI on legal help in Indian languages. A legal product with no Indian case law connector is a gap Anthropic can close with one partner deal.
Questions people are asking
What are Claude legal artifacts? They are ready-made tools inside Claude for Legal, added on 28 October 2026, for jobs like contract review, cite checks, regulatory scans and tracking open matters. Each opens with an example, and you add your own documents.
How much does Claude for Legal cost? It has no separate price. Artifacts in general open on every Claude plan, but connectors and saved storage need Pro, Max, Team or Enterprise.
Does Claude's cite check find fake cases? It flags quotes and claims that don't match the sources you give it. To check whether a case exists, use a case law connector such as CourtListener or Westlaw as well.
Can I share a Claude legal artifact with a client? You can invite up to 50 outside people, but the parts that use Claude or your connectors don't work for them.
Is Claude for Legal available in India? Yes, Claude plans are sold in India. But the gallery's named local tools are all for the US, and we found no Indian legal database among its published connectors.
Sources
- Anthropic, Introducing artifacts for legal work, 28 Oct 2026:claude.com/resources/…
- Claude on X, legal artifacts announcement, 28 Oct 2026:x.com/claudeai/…
- Anthropic, Claude for Legal:claude.com/solutions/…
- Claude Help Center, What are artifacts and how do I use them:support.claude.com/en/…
- Claude Help Center, Invite people outside your organization to an artifact, 23 Sep 2026:support.claude.com/en/…
- Claude Help Center, Share artifacts, 23 Sep 2026:support.claude.com/en/…
- Claude Help Center, Using Claude for legal work: privilege, confidentiality and configuration, 26 Aug 2026:support.claude.com/en/…
- Anthropic, CourtListener connector:claude.com/connectors/…
- Free Law Project, CourtListener is now available inside Claude, 12 May 2026:free.law/2026/…
- LawSites, Descrybe and Claude for the legal industry, 14 May 2026:lawnext.com/2026/…
- Damien Charlotin, AI Hallucination Cases database, updated 29 Oct 2026:damiencharlotin.com/hallucinations
- Reason (Volokh Conspiracy), 'Instead, Claude just made up more stuff', 19 May 2026:reason.com/volokh/…
- Fortune, Anthropic's legal plug-in and Claude for Legal, 12 May 2026:fortune.com/2026/…
- ABA Journal, Anthropic launches Claude for Legal, 14 May 2026:abajournal.com/news/…
- LawSites, Bob Ambrogi, Microsoft launches Copilot for Legal, 15 Oct 2026:lawnext.com/2026/…
- Latham & Watkins, Latham advises Harvey on financing round, Sep 2026:lw.com/en/…
- Legora, Legora raises $550 million Series E at $8.8 billion valuation, 21 Oct 2026:legora.com/newsroom/…
- Reuters (via Yahoo Finance), Anthropic's AI tools deepen selloff, 3 Feb 2026:finance.yahoo.com/news/…
- Intelligence by Intent, Steve Smith, Claude's new legal artifacts, 28 Oct 2026:intelligencebyintent.com/insights/…
- Business Today, Anthropic opens new office in Bengaluru, 16 Feb 2026:businesstoday.in/technology/…
- LiveLaw, Supreme Court publishes draft regulations on AI use in the judiciary, 3 Jun 2026:livelaw.in/top-stories/…
- LawSathi, Supreme Court releases draft rules for AI in courts, 15 Jun 2026:lawsathi.in/blog/…
- SupremeToday AI, BCI on lawyer accountability for AI hallucinations, 4 Jun 2026:supremetoday.ai/bci-rules-on-lawyer-accountability-for-ai-hallucinations-20260604091
- Bar & Bench, Shardul Amarchand Mangaldas announces partnership with Harvey:barandbench.com/news/…
- Bar & Bench, Cyril Amarchand Mangaldas partners with Legora, Jul 2025:barandbench.com/news/…
