Harvey Raises $550M at $15.5B Valuation to Build Its Own Legal AI Models

Harvey, the legal AI software company, has raised $550 million in new funding at a valuation of approximately $15.5 billion, more than tripling its valuation from $5 billion around 14 months ago.

The round was co-led by Lightspeed Venture Partners and Diffusion, a new investment firm founded by Kris Fredrickson, a longtime Harvey backer and former Coatue Management investor.

Sapphire Ventures and Whale Rock Capital Management also participated, alongside existing investors including Sequoia, Kleiner Perkins, Andreessen Horowitz and Goldman Sachs Alternatives.

The new funding comes as Harvey makes a significant strategic shift: building more of its own AI models rather than relying entirely on models supplied by companies such as OpenAI, Anthropic and Google.

Building its own legal AI models

Founded in 2022 by Winston Weinberg and Gabriel Pereyra, Harvey initially built its platform by routing legal workflows through foundation models from major AI companies.

Weinberg previously worked as a securities litigator at O'Melveny & Myers, while Pereyra was a research scientist at Meta and Google DeepMind.

In August, Harvey launched Tenet, its first proprietary model. The model was post-trained on Moonshot AI's Kimi K3 open-weight model with support from Fireworks AI.

The strategy gives Harvey greater control over the models powering sensitive legal workflows. Instead of relying on third-party models for every request, the company can increasingly tune models around the specific requirements of legal work.

Harvey CEO Winston Weinberg argues that post-training models will become an important capability for software companies competing in the AI era.

Acquiring AI security expertise

Alongside the funding announcement, Harvey also acquired Guardrails AI, an AI agent security startup focused on testing and evaluating autonomous AI systems.

The acquisition is Harvey's fourth of 2026, following Hexus in January, the Lume AI team in March and Benchmark in July.

Financial terms of the Guardrails deal were not disclosed. Harvey says these acquisitions are primarily focused on bringing engineering talent into the company rather than simply acquiring standalone products.

The Guardrails team adds expertise around evaluating what AI agents might do before they are allowed to interact with sensitive client information or systems.

Harvey's growing legal AI footprint

Harvey now serves more than 3,000 organisations, up from around 1,300 in March.

The company says annual recurring revenue has surpassed $400 million, while approximately 80% of Am Law 100 law firms use its platform.

Its customer base also includes five Fortune 10 companies and major organisations such as Latham & Watkins and Microsoft's in-house legal team.

This growth is taking place as Harvey's former model suppliers increasingly become competitors.

Anthropic has introduced legal-focused capabilities for Claude, while OpenAI has partnered directly with law firms to customise ChatGPT for legal work.

Harvey is therefore attempting to control more of its technology stack as competition increases.

A race to own the legal AI layer

Harvey operates in an increasingly crowded legal AI market, with companies such as Legora also expanding rapidly.

The global legal AI software market is projected to grow from approximately $5.21 billion in 2026 to $40.94 billion by 2034, according to Fortune Business Insights.

For Harvey, the next stage is about more than simply applying existing AI models to legal workflows. The company is attempting to build proprietary intelligence, acquire specialised AI talent and maintain control over the technology powering its platform.

The $15.5 billion valuation reflects investor confidence in that strategy. The challenge will be proving that owning more of the AI model layer can translate into a durable advantage as foundation-model companies and specialised legal AI competitors continue to advance.