Anthropic & Google Secure Harvey as an AI Tool User, Marking a Strategic Triumph

Anthropic & Google Secure Harvey as an AI Tool User, Marking a Strategic Triumph

Preface

In a significant development in the AI space, popular legal AI tool Harvey has announced its decision to incorporate leading foundational models from Anthropic and Google, in addition to those from OpenAI. This move, detailed in a recent blog post, marks a notable shift for Harvey, as it was initially a highly successful portfolio company backed by the OpenAI Startup Fund. While Harvey emphasizes it is not severing ties with OpenAI but rather expanding its model base, this evolution represents a substantial opportunity for its competitors like Anthropic and Google.

Lazy bag

Harvey, originally backed by OpenAI, is now embracing more AI models from Anthropic and Google, fueling competition and diversification.

Main Body

Harvey's strategic decision to incorporate AI models from Anthropic and Google marks a turning point in its trajectory. Established as one of the pioneering startups funded by the OpenAI Startup Fund, Harvey was distinguished by its early adoption and evolution using OpenAI's AI technologies. However, as the AI landscape evolved, Harvey recognized the potential of broadening its horizons through the integration of diverse AI models.

The OpenAI Startup Fund initially backed Harvey in December 2022 when Sam Altman, then CEO of OpenAI, was at its helm. Harvey, alongside other startups like Descript, Mem, and Speak, became a testament to OpenAI's commitment to fostering AI-driven innovations. Fast forward to February, Harvey, buoyed by its substantial growth, reached a $3-billion valuation, receiving substantial investment through a $300 million Series D funding led by Sequoia. This surge in capital, with contributions from OpenAI Fund and investors like Coatue and Kleiner Perkins, underscored its meteoric rise.

Intriguingly, Google had already shown interest in Harvey's potential. Its venture arm, GV, spearheaded a $100 million Series C in July 2024, which also saw participation from the OpenAI Fund. Despite this investment, Harvey initially hesitated to immediately adopt Google's AI models. Instead, it waited, conducting thorough internal evaluations.

This methodical approach led to the emergence of Harvey's internal benchmark, BigLaw, which demonstrated that various foundation models showed proficiency in a range of legal tasks. What stood out was the variance in performance; some models excelled at certain tasks but underperformed in others. Specifically, Google's Gemini 2.5 Pro demonstrated excellence in legal drafting but struggled with pre-trial tasks like formulating oral arguments, primarily due to its limited grasp of complex legal concepts like evidentiary rules. On the contrary, OpenAI's models displayed strong performance in these pre-trial tasks, with Anthropic's Claude 3.7 Sonnet trailing closely.

Aware of these insights, Harvey took the strategic decision not to develop its own models but to incorporate high-performance models from multiple providers. This decision allows Harvey to refine these models specifically for the legal sector, optimizing their capabilities for distinct legal tasks. Additionally, deploying a variety of AI models enables Harvey to innovate more efficiently in developing AI agents tailored for legal applications.

In a noteworthy move, Harvey has also committed to enhancing transparency by joining the ranks of companies publishing public leaderboards for model performance. Harvey plans to publish detailed insights, presenting legal task benchmark performances not just as single scores but as nuanced evaluations contributed by top legal professionals.

This development not only signals Harvey's adoption of competitive models but also creates a challenge for its backers. Companies like OpenAI and Google now face the imperative of continuously outperforming new entrants in this evolving landscape. Despite the expanding complexities in AI benchmarking, OpenAI still enjoys considerable esteem in the industry.

Harvey CEO Winston Weinberg expressed optimism on this strategic expansion, recognizing the invaluable role OpenAI plays as both investor and collaborator. “We are incredibly fortunate to have OpenAI as an investor in Harvey and key collaborator in our product,” he stated, emphasizing Harvey’s ongoing commitment to enhancing global customer experiences by diversifying technology partnerships.

Key Insights Table

AspectDescription
Key Fact 1Harvey integrates Anthropic and Google's models along with OpenAI's to diversify AI tool capabilities.
Key Fact 2Internal benchmarking shows varying strengths among different models for specific legal tasks.
Last edited at:2025/5/13

Mr. W

ZNews full-time writer