Harvey raises $550M more to develop AI tools for legal teams
SiliconANGLE Maria Deutscher ● Covered by 3 sources
Harvey just raised $550M at a $15.5B valuation to build AI tools for lawyers. It’s leaning hard into custom models, not just using everyone else’s.
Based on reporting by SiliconANGLE, Maria Deutscher — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Harvey keeps showing that legal AI is still a very lucrative business. Six months after its last nine-figure round, the company has pulled in another $550 million, this time at a $15.5 billion valuation. Diffusion and Lightspeed Venture Partners led the deal, with more than a dozen other investors joining in, including Sequoia, Kleiner Perkins and Goldman Sachs.
The pitch is simple enough: Harvey sells a cloud platform that helps law firms and in-house legal teams automate work that used to eat up attorney time. The company says 80% of the 100 highest-ranked law firms in the U.S. use its software, and it counts half the Fortune 10 as customers. That gives Harvey a footprint most startup vendors would kill for, especially in a field where trust and switching costs matter.
Its platform centers on a document repository called Vault, which can store up to 100,000 documents. Users can search it with AI to pull out patterns, clause language and other useful material, whether they’re checking supplier contracts against a new regulation or digging through old customer agreements for language to reuse in a new deal. Harvey’s system also reaches outside the company archive for precedents and legislative clauses.
The company has also been pushing beyond search and retrieval. Earlier this year it introduced AI agents for more complex work, such as combing through due diligence documents and flagging possible issues, while asking lawyers for help when a task gets ambiguous. And just a few days ago it launched Tenet, its first custom large language model, built from Kimi K3, trained on legal documents and wrapped in a custom harness meant to improve output quality.
Harvey says Tenet is 20% better than Kimi K3 on some contract-processing tasks, and that it beats Fable 5 and GPT-5 Sol in multiple areas. The company also says it plans to add more computing infrastructure and focus on “new generalist models.” That last part matters as much as the funding: custom models are expensive, but if Harvey can keep moving work off external systems, the economics should get better over time.
My take — AI-written commentary, not fact-checked reporting
This is the sensible kind of AI arms race: boring, pricey, and aimed squarely at workflow, not swagger. Legal teams do not buy vibes; they buy less drudgery and fewer missed clauses. The real tell is Harvey pouring money into its own models, because the smartest AI companies are already trying to own the stack before the rents go up.
Read more about this at: SiliconANGLE
