Harvey AI
AI platform that helps lawyers draft, review, and research using custom legal agents.
On the record
Checkpoint: EU AI Act compliance obligations take effect for legal AI tools
The call
Harvey grew through Anthropic's legal plugin launch and Kirkland's in-house AI bet still looms as the test of whether BigLaw keeps buying or starts building.
Harvey builds artificial intelligence software that helps lawyers at major law firms do research and write documents faster. It is somewhat hard to copy because Harvey has spent years training its system on legal documents, built custom tools for each client, and hired former lawyers who know how to sell to law firms. However, big companies like Thomson Reuters and LexisNexis already own huge libraries of court decisions and legal publications, and they are now adding AI assistants to their existing products. The biggest question is whether Harvey can keep its legal information as accurate and current as these larger competitors, or whether major law firms will just build their own AI systems instead of paying Harvey.
Why
01
AI pressure: Anthropic's Claude for Legal and Claude Cowork. Anthropic's legal plugins launched inside Harvey's own customer base, and Harvey's ARR rose through the launch rather than falling. Could AI do the core job? Unlikely.
02
Bigger players: Thomson Reuters (CoCounsel) and RELX/LexisNexis (Lexis+ AI). Both own primary-law databases and have shipped AI copilots, but have moved slower than Harvey in landing BigLaw seats so far. Could a giant copy it? Probably.
03
What protects it. Liability-grade accuracy (0.2% hallucination rate on BigLaw Bench), 25,000+ custom firm agents, and an ex-lawyer sales force built into every major firm's workflow.
What would change this call
Whether two or more AmLaw 20 firms announce internal-build replacements for Harvey by end of 2027. Would show BigLaw self-disintermediation outpacing Harvey's lock-in, echoing Kirkland's $500M build.
Whether Harvey's ARR growth rate falls below roughly 50% YoY in any quarter through 2027. Would show incumbents or model-native tools eroding Harvey's growth, not just competing at the margin.
Whether a named AmLaw 100 firm publicly shifts primary-law research workload to CoCounsel, Lexis+ AI, or Legora. Would show incumbents converting their primary-law data advantage into actual customer switching.
Whether Harvey's ARR growth rate falls below roughly 50% YoY in any quarter through 2027. Would show incumbents or model-native tools eroding Harvey's growth, not just competing at the margin.
Whether a named AmLaw 100 firm publicly shifts primary-law research workload to CoCounsel, Lexis+ AI, or Legora. Would show incumbents converting their primary-law data advantage into actual customer switching.
Key numbers
- ARR hit $300M, up from $195M, in five months
- Roughly 50% of Am Law 100 are paying customers
- Kirkland & Ellis committed $500M to build AI in-house
Named risks
- Big Law self-build (e.g. Kirkland's $500M AI investment) could disintermediate Harvey from top firms.
- Industry-wide AI hallucination liability incidents (1,490+ court cases by mid-2026, sanctions up to $145,000+) could hit Harvey's reputation.
- Frontier labs like Anthropic and OpenAI could absorb legal AI workflows directly, compressing Harvey's differentiation over time.
- Incumbents (Thomson Reuters, RELX) own primary-law data Harvey lacks, a structural grounding gap critics could exploit.
- Revenue figures are third-party estimates, not audited disclosures, creating uncertainty about true financial health.
Rivals, as read in July 2026
| Rival | Direction | Evidence |
|---|---|---|
| Anthropic (Claude for Legal) | contesting | Launched Feb 3, 2026 legal plugin and May 2026 Claude for Legal; integrates with rather than replaces Harvey |
| Thomson Reuters (CoCounsel) | losing to | Shares dropped 18% on Anthropic launch Feb 2026; slower AI shipping cited versus Harvey |
| RELX/LexisNexis (Lexis+ AI) | losing to | Shares dropped 14% on Anthropic Feb 3, 2026 launch; owns primary-law data Harvey lacks |
| Legora | contesting | Stockholm rival valued ~$5.6B, raised $265M+ as of Jan 26, 2026, strong in Europe |
| Kirkland & Ellis (in-house build) | gaining on | Announced $500M internal AI investment June 8, 2026, signaling possible self-build over buying Harvey |
Questions people ask
- Is Harvey AI profitable in 2026?
- No public profitability figures exist. Harvey is a private, VC-backed company reporting rapid revenue growth (estimated $300M ARR by May 2026, per Sacra), but has not disclosed profitability. It closed a $200M round at an $11B valuation on March 25, 2026, suggesting continued growth-stage cash burn rather than confirmed profit.
- Can AI replace Harvey AI's own product?
- Not yet demonstrated. Anthropic's Claude for Legal launched May 2026 and integrates with Harvey rather than replacing it; Harvey's ARR grew from $195M to about $300M through that period. Gartner analysts said Anthropic's plugin was not a commercial threat to specialist vendors like Harvey as of July 8, 2026.
- Who is Harvey AI's biggest competitor?
- Thomson Reuters (CoCounsel/Westlaw) and RELX/LexisNexis (Lexis+ AI) are the main incumbent rivals, owning primary-law databases; Legora, a Stockholm-based startup valued around $5.6B as of January 2026, is Harvey's closest direct AI-native peer competitor.
- Who owns Harvey AI?
- Harvey AI Inc. is privately held and venture-backed, founded in 2022 by CEO Winston Weinberg and Gabriel Pereyra. Investors include Sequoia (co-led three rounds), Andreessen Horowitz, Kleiner Perkins, Coatue, GIC, and OpenAI Startup Fund, which seeded the company in November 2022; no single owner or parent company exists.
- Is Harvey AI losing customers to Big Law self-build efforts?
- No confirmed customer losses have been reported. Kirkland & Ellis announced a $500M internal AI investment on June 8, 2026, signaling possible self-build risk, but Harvey's logo base and ARR (growing to about $300M by May 2026) show no visible erosion as of mid-2026.
- Does Harvey AI hallucinate less than other legal AI tools?
- Harvey's own BigLaw Bench benchmark, published October 7, 2024, claimed its Assistant model hallucinates about 0.2% of claims versus 0.7 to 1.9% for raw foundation models (vendor-claimed figure). Independent critics have disputed legal AI accuracy broadly, citing higher hallucination rates industry-wide.
Timeline
- 2026-07-17Ninth Circuit issues binding sanctions precedent for AI hallucinations in filings
- 2026-06-08Kirkland & Ellis announces $500M internal AI investment
- 2026-05-30Anthropic launches Claude for Legal with 20+ connectors, integrates with Harvey
- 2026-03-25Harvey raises $200M Series at $11B valuation, co-led by GIC and Sequoia
- 2026-03-01Harvey launches custom agent-building tool; 25,000+ agents built to date
- 2026-02-03Anthropic launches Claude Cowork legal plugin, triggers 13 to 18% drop in TR, RELX, Wolters Kluwer shares
- 2026-01-26Sacra reports Harvey at $195M ARR, 290% YoY growth
- 2025-12-31Harvey closes $160M Series F at approximately $8B valuation, led by a16z
- 2025-06-25Harvey raises $300M Series E at approximately $5B valuation
- 2024-10-07Harvey publishes BigLaw Bench hallucination benchmark results
Revision history
20 July 2026
First assessed. Defensible for now. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources
- Harvey revenue, valuation & funding
- Harvey Raises at $11 Billion Valuation (2026-03-25)
- Legal AI startup Harvey valued at $11B (2026-03-25)
- BigLaw Bench: Hallucinations (2024-10-07)
- Two Attorneys Just Lost Six Months of Practice Rights (2026-07-17)
- One Month On, the Market Fallout... (2026-03-05)
- Anthropic's Legal Plug-Ins Signal a New Era (2026-05-30)
- Harvey 2026 Company Profile
- What Legal Teams Can Learn from Sullivan & Cromwell's AI Hallucination Crisis (2026-07-16)
- Harvey Hits $8B Valuation in Third 2025 Funding Blitz
- Bell Gully Bought Harvey Firm-Wide (2026-06-09)
- Harvey at $195M ARR (2026-01-26)