Legal-software startup Harvey has introduced its first proprietary AI model, Tenet, aiming to reduce reliance on third-party providers and improve performance for complex legal tasks.

Key Points

• Harvey Tenet is a bespoke AI model trained on legal reasoning data created by attorneys and built upon the open-source Kimi K3 architecture.
• The company aims to lower operational costs by reducing dependency on expensive third-party models from providers like OpenAI and Anthropic.
• The rollout is part of the broader "Harvey II" update, which includes a new "Memory" feature for storing user-specific work preferences.
• Harvey plans to eventually allow law firms to use Tenet as a foundation to train their own internal models on proprietary firm data.
• The move comes as major AI developers, including OpenAI and Anthropic, increasingly target the legal technology market with their own specialized tools.

Why it Matters

Developing a proprietary model allows Harvey to secure its profit margins and gain independence from the very companies that are becoming its direct competitors. This strategic shift positions Harvey to evolve from a software provider into a specialized infrastructure partner that helps law firms institutionalize their internal expertise.