Industry Guide: Legal
How to Use Claude in Legal: A Guide for Lawyers, In-House Counsel, and Legal Ops
Faster drafts, same privilege standard.
This guide is for litigators, transactional lawyers, in-house counsel, legal operations, and legal technology teams who want to put Anthropic's Claude to work on the contracts, research, and case files that move legal work forward. It covers Anthropic's own legal releases and what named legal companies said about using Claude in Anthropic's announcements and customer pages.
Knowing how to use Claude in legal means focusing it on document-heavy, privilege-sensitive work: reviewing contracts against a playbook, drafting demand letters and correspondence from case files, triaging intake and conflicts, researching case law and statutes, tracking regulatory changes across jurisdictions, and summarizing discovery materials for litigation teams. Claude drafts, extracts, and flags inconsistencies. Attorneys and qualified legal professionals own every judgment call on privilege, strategy, and what gets filed, and firms keep AI-assisted work inside their existing review and confidentiality controls.
Published
Free lesson · no account neededThe Iteration LoopMaster the Essentials6 minStart the free lessonKey Takeaways
- The strongest legal uses are contract review, legal research, litigation support, intake and triage, and regulatory tracking.
- Anthropic's legal releases include a solutions page for legal teams and a gallery of ready-to-use legal artifacts built with law firms and legal technology companies.
- Anthropic's customer pages describe how Eve Legal, EvenUp, Spellbook, GC AI, Harvey, and Steno use Claude in their own products and workflows.
- Every AI-drafted legal document needs a human reviewer, and privileged or confidential material requires an approved, access-controlled deployment.
- Attorney judgment on strategy, privilege, and what gets filed with a court or counterparty stays with qualified legal professionals.
Copy and Paste
Prompts You Can Use Today
Contract review against a playbook
You are reviewing a vendor services agreement against our standard playbook pasted below. Flag every clause that deviates from the playbook position, including limitation of liability, indemnification, termination, and data protection terms, cite the clause number for each flag, suggest a fallback position consistent with the playbook, and list anything missing that the playbook requires. Playbook and agreement: [paste].
Playbook deviation review is one of the clearest wins because the standard to check against is explicit.
Demand letter draft from case file notes
Draft a demand letter using only the case file summary, medical records excerpt, and liability notes pasted below. State the facts exactly as described, reference the supporting document after each factual claim, do not assert a damages figure beyond what the notes support, and flag any gap in the record that should be filled before this goes out. Case file materials: [paste].
Grounding every factual sentence in a cited document is what makes the draft safe to review quickly.
Legal intake and conflict triage
Review the new matter intake form below and extract the parties involved, the matter type, the jurisdiction, the requested deadline, and any party name that should be run through a conflict check. List anything the intake form is missing that our standard checklist requires, and flag any detail that looks like it needs partner attention before the matter is opened. Intake form: [paste].
This speeds up triage without letting the model decide whether a conflict actually exists.
Regulatory horizon scan
Summarize the regulatory updates below that are relevant to our compliance team: the effective date, the jurisdiction, what changed compared to the prior rule, and which of our existing policies or contract templates might need review as a result. Cite the specific update for each point and do not speculate about updates not included here. Regulatory updates: [paste].
Keeping the scan tied to the pasted updates avoids the model guessing at rules it has not seen.
Discovery document summary for litigation
Summarize the attached deposition transcript excerpt for the litigation team: each topic discussed, the witness's stated position on each topic, any point where the witness's testimony is inconsistent with the exhibit referenced, and the page and line citation for every point you make. Do not characterize credibility or strategy. Deposition excerpt and exhibits: [paste].
Page and line citations are what let an attorney verify the summary in minutes instead of rereading the whole transcript.
From Anthropic
What Anthropic offers for Legal
Claude for legal teams
Anthropic's solutions page for legal teams describes tabular contract review, regulatory tracking, legal intake, jurisdiction surveys, playbook deviation review, and matter tracking built into Claude, along with integrations for Word and Outlook and connectors to iManage, NetDocuments, Ironclad, and Bloomberg Law, plus privacy commitments for privileged work on Team and Enterprise plans. The page also notes that a Harvey executive quoted there said Claude Opus 4.7 scored 90.9% on Harvey's BigLaw Bench evaluation.
Introducing artifacts for legal work with Claude
Anthropic's October 2026 article introduces a gallery of ready-to-use legal artifacts built with law firms and legal technology companies including Goodwin, Harvey, Foley and Lardner, and PwC, covering a cite check, a jurisdiction survey matrix, a matter board, a playbook deviation board, a regulatory horizon scan, and tabular review, alongside a Legal Ambassadors Program and access-to-justice tools built with nonprofits such as Zealous, Mobile Pathways, Rising for Justice, and The Access Project. Anthropic says the release adds more than 60 legal connectors.
· Source
In Practice
How Legal Teams Use Claude
According to Anthropic's published customer story
Eve Legal
Eve Legal is an AI-native platform for plaintiff-side law firms, built on Anthropic's Claude to process case files and draft demand letters; the company also uses Claude Code for its own engineering work. Eve reports 60 days faster settlement times, says it can increase settlement values by up to 30%, and that lawyers can take on up to 30 additional cases per lawyer per year while the platform processes 12.5 million documents a month. More than 1,700 firms run on Eve, with 500 of them added this quarter alone, and the company says it has cut engineering triage from three hours to minutes across around 420 test suites, automating 80% of the work, so that case life cycles run 50% faster.
According to Anthropic's published customer story
EvenUp
EvenUp built its AI Drafts product for personal-injury cases on Claude. The company says the tool cuts document drafting from 15 hours to 15 minutes, down from 8 to 15 hours of skilled work per case, and reports a 300% increase in settlement offers. EvenUp states it has resolved more than 200,000 cases and secured more than $10 billion in damages, that revenue has grown 400% in one year, and that more than 200 employees use Claude Enterprise internally, with accuracy on past medical expense totals increased by 14%.
According to Anthropic's published customer story
Spellbook
Spellbook builds contract drafting and review tooling for law firms and in-house teams on Claude. The company reports 530,000 contract reviews per month and says it delivers 10x faster agreement turnaround time, taking a review from roughly 10 hours of lawyer work to about one. Spellbook says its product handles more than 700,000 chat messages each month, has grown to 5,000 customers across 80 countries, and runs roughly 15 Claude agent configurations.
According to Anthropic's published customer story
GC AI
GC AI built an in-house legal workflow platform used by more than 1,500 companies. Based on a survey of more than 100 active customers, GC AI reports 14 hours saved per week, a 14% reduction in outside counsel spend, and 21% greater accuracy on legal tasks compared to general-purpose AI tools. The company also says its users can absorb a 30% increase in deal volume.
According to Anthropic's published customer story
Harvey
Harvey is a legal AI platform used by law firms and Fortune 500 companies. Anthropic's customer page for Harvey describes a deployment time of under 1 month and one of the highest performance scores on Harvey's own BigLaw Bench evaluation.
According to Anthropic's published customer story
Steno
Steno is a legal support and court-reporting company that built its Transcript Genius product using Claude 3 Opus's 200K-token context window; more than 1,200 law firms use Steno's platform.
Illustrative walkthrough
A contract review sprint before signing
- An associate loads the draft agreement, the firm's playbook position for each clause type, and the house style guide into a shared workspace.
- Claude flags every clause that departs from the playbook position and proposes a fallback consistent with that position, citing the clause number for each flag.
- The reviewing attorney checks each flagged clause against the actual deal terms and resolves anything the playbook does not clearly cover.
- A second reviewer confirms defined terms and dates are consistent across every exhibit and schedule.
- The final version moves to signature through the firm's normal document control process.
Core Workflows
Claude in Legal Use Cases From Intake to Resolution
These are the areas where AI for legal teams saves the most attorney and staff time while keeping privilege and accuracy intact.
Contract review and drafting
AI for contract review flags playbook deviations, drafts clauses from precedent language, and extracts key terms for a repository, with every suggestion tied back to the governing clause.
Legal research
AI legal research pulls together case law, statutes, and secondary sources into a structured starting point that a lawyer still verifies before relying on it.
Litigation support
Summarizing depositions, organizing exhibits, and drafting discovery responses from the underlying record, with page and line citations a litigator can check.
Legal intake and triage
New-matter intake, conflict-check flagging, and routing requests to the right team, so legal operations spends less time on administrative sorting.
Regulatory tracking
Monitoring rule changes across jurisdictions and mapping each one to the policies or templates it affects, ahead of a compliance deadline.
Compliance workflows
Vendor NDA triage, policy question answering, and templated responses to routine requests, configured to a legal team's own playbook.
What Legal Companies Report
How Law Firms and Legal Teams Use AI Today
How is AI used in law firms? Anthropic's customer pages describe specific products built on Claude. Eve Legal, an AI-native platform for plaintiff-side law firms, uses Claude to process case files and draft demand letters, and relies on Claude Code for its own engineering work. EvenUp built its AI Drafts product for personal-injury cases on Claude, aimed at the medical-record review and damages narrative work that precedes a settlement demand.
On the transactional side, Spellbook built contract drafting and review tooling for law firms and in-house teams, and GC AI built an in-house legal workflow platform used by more than 1,500 companies, according to Anthropic's customer page for GC AI. Harvey, a legal AI platform used by law firms and Fortune 500 companies, and Steno, a legal support and court-reporting company, are both described on Anthropic's customer pages as well.
Across AI for law firms and AI for in-house legal teams, and AI in the legal industry more broadly, the pattern on these pages is consistent: generative AI in legal tends to start with document-heavy, repetitive work, with contract review, drafting, research, and intake replacing hours of manual effort while the people who own legal judgment keep reviewing the output.
Anthropic's Releases
The Legal Tools Anthropic Has Built
Anthropic's solutions page for legal teams lays out the areas it has built for: tabular contract review, regulatory tracking, legal intake, jurisdiction surveys, playbook deviation review, and matter tracking, with integration into Word and Outlook and connectors to legal systems such as iManage, NetDocuments, Ironclad, and Bloomberg Law. The page also sets out privacy commitments for privileged work on its Team and Enterprise plans.
In late October 2026, Anthropic published an article introducing a set of ready-to-use legal artifacts built together with law firms and legal technology companies, including Goodwin, Harvey, Foley and Lardner, and PwC. The gallery covers a cite check, a jurisdiction survey matrix, a matter board, a playbook deviation board, a regulatory horizon scan, and tabular review, and Anthropic says the release adds more than 60 legal connectors in total.
The same article describes a Legal Ambassadors Program and access-to-justice tools built with nonprofits including Zealous, Mobile Pathways, Rising for Justice, and The Access Project. Anthropic also lists a legal plugin on its marketplace for in-house teams, built around slash commands for contract review, NDA triage, vendor checks, briefings, and templated responses, configurable to a firm's own playbook.
Writing in Practice
AI for Contract Review Done Well
The pattern that holds up in contract review is a named playbook and hard references. Give Claude the agreement, the exact playbook position for each clause type, and ask it to flag every deviation by clause number rather than rewrite the document outright. A review built that way can be checked clause by clause, which is what a supervising attorney actually needs.
Consistency checking across a document set is an underrated use. A deal with dozens of ancillary agreements repeats the same defined terms and dates, and a mismatch in one exhibit creates real risk. Claude can compare documents and list every place they disagree, which turns a full reread into a short, focused list.
Keep the division of labor explicit. Claude drafts, extracts, and flags. The reviewing attorney owns accuracy and the final clause language, and the partner or general counsel owns what gets signed or filed. Writing that division into your workflow makes AI assistance easier to defend if a client or a court ever asks how a document was produced.
Choosing Tools
Claude vs ChatGPT for Legal Work
The Claude vs ChatGPT for legal comparison is most useful run on your own documents: a contract already negotiated to close, a research memo already checked by a partner, and a demand letter already sent. Compare accuracy against the known-correct version, citation behavior, and how each tool handles a gap in the source material instead of guessing.
In legal practice the decision usually comes down to governance rather than a single benchmark: which deployment your firm's or company's IT and risk teams have approved, how client confidentiality and privilege are protected, where data is stored and processed, and whether the tool connects to the document management and research systems you already use.
Is it safe to use Claude for privileged legal work? Privilege and confidentiality obligations do not change because a draft started with AI. Anthropic's solutions page for legal teams describes privacy commitments on its Team and Enterprise plans, but whether a given deployment meets your firm's or client's requirements is a question for your own risk, security, and ethics counsel to confirm.
Getting Started
A Safe First Month for a Legal Team
Start with closed matters
Test prompts on a matter that is already resolved, and compare the AI output against what was actually filed or sent.
Load your playbook and templates
Put your contract playbook, templates, and house style into a project so drafts and reviews match your existing standards.
Pick one document type
Choose one high-volume task, such as NDA review or intake triage, and measure the change in review time.
Define the privilege boundary
Agree with risk and IT which work can touch privileged or client-confidential material and on which approved deployment.
Limits
Where Qualified Legal Professionals Must Decide
Claude can misstate a fact, invent a case citation, or drift into a legal conclusion the record does not support. Every AI-assisted filing, letter, or advice needs a qualified reviewer, and every citation to case law or a statute should be checked against the actual source before it goes anywhere near a court or a client.
Does using AI risk attorney-client privilege? The risk is not the drafting tool itself; it is where confidential material is sent and stored, and whether that deployment is one your firm or client has actually approved. Strategy calls, privilege determinations, and what ultimately gets filed or signed stay with the attorney of record.
Treat case files, medical records, and unredacted discovery material as sensitive by default. Use only an approved deployment, follow your firm's data handling rules, and keep the final record of what was filed or sent inside your normal document management and billing systems.
Build Reliable AI Habits for Legal Work
Getting trustworthy drafts from AI is a practiced skill: grounding, citation discipline, and review. The free Nightschool AI curriculum is hands-on from the first lesson.
Frequently Asked Questions
How is AI used in law firms?
Mostly for document-heavy, repetitive work: contract review, legal research, drafting, intake triage, and litigation support such as discovery summaries. Anthropic's customer pages describe Eve Legal, EvenUp, Spellbook, GC AI, Harvey, and Steno building products on Claude for exactly these tasks. Attorneys keep reviewing the output and own every legal judgment.
Can AI review contracts?
Yes, for flagging deviations from a playbook, extracting defined terms and dates, and checking consistency across a document set. Spellbook and the legal plugin on Anthropic's marketplace both target contract review specifically. A lawyer still has to decide whether a flagged deviation is acceptable for this deal.
Is it safe to use Claude for privileged legal work?
Privilege obligations do not change because a draft started with AI. Anthropic's solutions page for legal teams describes privacy commitments on its Team and Enterprise plans, but whether a specific deployment meets your firm's or client's requirements is a question for your risk, security, and ethics counsel, not a general yes.
How much time can AI save on legal drafting?
It depends heavily on the document type and how clear your source material is, so measure it on your own matters. Time savings come from drafting against a playbook or case file rather than a blank page. Review and sign-off time does not disappear, and it should not.
What are the biggest AI use cases in legal?
The most common are contract review and drafting, legal research, litigation support such as deposition summaries, legal intake and conflict triage, and regulatory tracking across jurisdictions. Each works best with a named source document and a human reviewer checking the output before it is relied on.
Can AI help with litigation support?
Yes, for summarizing depositions and discovery materials, organizing exhibits, and drafting routine filings from the underlying record, with page and line citations so a litigator can check the summary quickly. Case strategy, credibility assessments, and what ultimately gets filed stay with the litigation team.
Does using AI risk attorney-client privilege?
The risk comes from where confidential material is sent and stored, not from drafting assistance itself. Use only a deployment your firm or client has actually approved, and keep privilege determinations and filing decisions with the attorney of record. Anthropic's solutions page describes privacy commitments on its Team and Enterprise plans.
How does Claude compare to ChatGPT for legal work?
Run the comparison on your own finished documents: a negotiated contract, a checked research memo, a sent demand letter. Compare accuracy, citation behavior, and handling of gaps in the source material. In practice, the deciding factors are usually governance and connectors to your existing legal systems, not a single benchmark.