Claude for Legal is Anthropic's offer for legal teams, launched in May 2026, and on 28 October 2026 it gained a gallery of ready-to-use artifacts for legal work, such as a cite checker and a matter board. Teams adapt them with their own data and connectors, then share them across the business.
Artifacts are interactive tools, like trackers and dashboards, that teams use and share inside Claude. Each one opens with an example, so you can see how the work comes together before adding your own data, and every answer or output links back to its source. The workflow Anthropic describes is simple: the legal professional brings their own documents and sources, Claude does the first pass, and a lawyer applies judgement to the results.
The six legal artifacts and what each one does
| Artifact | What it does, per Anthropic |
|---|---|
| Cite check | 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 |
| Jurisdiction survey matrix | Compares one legal topic, such as data breach notice rules, across states or countries in a single table, with links to the underlying law |
| Matter board | Turns engagement letters, emails or spreadsheets into one board of open matters, showing owner, next step and next deadline |
| Playbook deviation board | Compares contracts with your standard terms, flags each clause that differs and sends it to the person who approves exceptions |
| Regulatory horizon scan | Maps upcoming regulations on a timeline based on your business and where it operates, ranked by impact, with links to sources |
| Tabular review | Builds a table from a set of contracts showing the terms you ask about, such as severance or non-competes, with links to the exact language |
The cite check and tabular review are the two I'd point a small in-house team at first, because both end with a link to the exact wording, so a lawyer can check every line rather than trust a summary.
Connectors, partner-built tools and sharing
Since launch, Anthropic says it's added more than 60 legal connectors, including Bloomberg Law, Filevine, Ivo, Legora and Patlytics. Connectors let artifacts draw on your playbooks or research sources without anyone uploading files by hand.
The gallery also carries artifacts built by firms and partners: an AI Governance Benchmark from Goodwin, Harvey Intake from Harvey, an Obligations Checker from Foley and Lardner, and a Subpoena Response Tool from PwC, with more coming from Legora and Orrick. Artifacts can be shared with colleagues across the business, or with clients and partners, using sharing settings that Anthropic says support privacy and ethical walls between teams. Anthropic also says it's funding public interest groups building access-to-justice artifacts, funding legal education programmes and starting a Legal Ambassadors Program.
If you're weighing Claude against other options for a small firm or business, my roundup of the best AI legal tools for small businesses sets them side by side.
Anthropic's view on privilege, as of June 2026
Anthropic has published an article on privilege and confidentiality. It says plainly that it's Anthropic's perspective on its own products and the public record as of June 2026, that it isn't legal advice, and that the right answer depends on jurisdiction, clients and matters. I'm summarising what Anthropic says, not giving a legal opinion.
- Anthropic argues the profession has been here before with email and cloud storage. It points to bar guidance such as ABA Formal Opinions 99-413 and 477R, which concluded lawyers may use those technologies with reasonable care in choosing and supervising the vendor, and to ABA Formal Opinion 512, from 2024, which applies the same duties to generative AI.
- It cites United States v. Heppner, a February 2026 case in the Southern District of New York, which held that a criminal defendant's own use of a consumer AI account, without his lawyers' involvement, wasn't privileged work product. Anthropic notes the reasoning turned on consumer privacy terms and the absence of attorney direction, and said nothing about lawyers using enterprise tools.
- It cites later non-appellate decisions that went the other way for AI-assisted litigation preparation, Warner v. Gilbarco, Morgan v. V2X, Tate Group Automotive v. Legacy Automotive Capital and Assini v. Hayward, which each held it was protected work product and that sharing inputs with an AI tool doesn't by itself waive protection. Anthropic adds that courts elsewhere may differ.
- It notes that work-product protection is generally tied to litigation, so for other work the focus is attorney-client privilege where it applies, plus contractual no-training and confidentiality commitments.
Anthropic's stated view is that Claude's commercial offerings can be configured by lawyers in a way that maintains privilege and confidentiality. That's the vendor's position, and it's one for your own counsel to test.
What Anthropic keeps, and who can read it
Under commercial terms, customer content from Claude for Work, the Claude Platform and the API isn't used to train models by default. Claude for Work administrators can set an organisation-wide retention period, and API customers get standard 30-day retention or, on eligible endpoints, Zero Data Retention.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
For its most capable models, Anthropic keeps prompts and outputs for 30 days in a governed store that's deleted automatically. Automated safety systems process that data, and only content they flag can reach a human reviewer, drawn from a small designated safeguards group, with two-person approval for regulated data categories. Customers with Access Transparency can pull a tamper-proof log of every human read. Customer-managed encryption keys are available to qualifying customers, which is about key custody, not switching off safety review. Running Claude through AWS Bedrock or Google Cloud Vertex AI changes where the data sits but doesn't disable safety monitoring either. For the wider security picture, my guide to Claude Security is the next read.
A risk checklist before your legal team starts
- Make sure you're on commercial terms, with a data processing agreement and the no-training commitment in place, not a consumer account.
- Decide whether sensitive matters need Zero Data Retention or an in-tenancy route through your own cloud.
- Set your retention period in Claude for Work before people start uploading matter documents.
- Keep a lawyer in the loop on every output, and verify it against primary sources.
- Document how and where your team uses AI on each matter.
- Follow your own client-consent and engagement-letter practices, which ABA Formal Opinion 512 addresses.
- Ask your own counsel to review Anthropic's privilege article against your jurisdiction and client base.
My verdict on Claude for Legal
The artifacts look most useful for the grinding first-pass work, like reviewing a contract set or checking citations, and every output links back to its source, which makes review faster. My advice stays on the workflow side: I wouldn't let anyone on the team use a personal consumer account for client work, given that's the risky case Anthropic itself points to. I'd check the commercial terms and retention settings first, keep a lawyer signing off every result, and treat Anthropic's privilege article as its view, not a ruling you can rely on.
If you're bringing Claude into a legal team and want the workflow and guardrails set up before anyone starts, that's the kind of thing I help with. I run AI training and workshops for teams, and help businesses with AI enablement, AI automation and GEO, which is about getting your business seen and cited in AI answers. Most of that work is with SaaS marketing teams, telecom companies and small businesses. Here's how we can work together.
Frequently asked questions
What is Claude for Legal?
Claude for Legal is Anthropic's offer for legal teams, launched in May 2026. On 28 October 2026 Anthropic added a gallery of ready-to-use legal artifacts, such as cite check and matter board, that teams adapt with their own data and connectors.
Does using Claude waive legal privilege?
Anthropic's article, written as of June 2026, says its view isn't legal advice and the answer depends on jurisdiction, clients and matters. It says Claude's commercial offerings can be configured to maintain privilege and confidentiality.
Does Anthropic train on legal teams' data?
Under commercial terms, customer content from Claude for Work, the Claude Platform and the API isn't used to train models by default.
Which legal connectors does Claude have?
Anthropic says it's added more than 60 legal connectors since launch, including Bloomberg Law, Filevine, Ivo, Legora and Patlytics.
