Harvey, a startup building what it describes as an AI-powered “copilot” for lawyers, has raised $100 million in a Series C round led by GV, Google’s corporate venture arm. The tranche, which also had participation from heavy-hitting angels and VCs OpenAI, Kleiner Perkins, Sequoia Capital, Elad Gil and SV Angel, brings Harvey’s total raised to […]
© 2024 TechCrunch. All rights reserved. For personal use only.
Harvey, a startup building what it describes as an AI-powered “copilot” for lawyers, has raised $100 million in a Series C round led by GV, Google’s corporate venture arm.
The tranche, which also had participation from heavy-hitting angels and VCs OpenAI, Kleiner Perkins, Sequoia Capital, Elad Gil and SV Angel, brings Harvey’s total raised to $206 million and values the firm at $1.5 billion.
In a post on Tuesday on Harvey’s official blog, co-founders Winston Weinberg and Gabriel Pereyra said that the bulk of the new capital will be put toward collecting and curating data to build and train “domain-specific” AI models while growing Harvey’s headcount and expanding its paid services to new geographies.
“This investment will enable Harvey to continue scaling and improving our AI-powered technology across business functions and geographies,” Weinberg and Pereyra said. “We will use this new capital to invest in the engineering, data and domain expertise that are fundamental to building AI-native systems that facilitate the most complex knowledge work. We will also deepen our partnerships with both cloud and model providers to integrate additional models into Harvey and broaden our training collaborations to continue improving model efficacy.”
Weinberg, a former securities and antitrust litigator at law firm O’Melveny & Myers, and Pereyra, previously a research scientist at DeepMind, Google Brain (another of Google’s AI groups) and Meta AI, launched San Francisco-based Harvey in 2022. Weinberg and Pereyra are roommates; Pereyra showed Weinberg OpenAI’s GPT-3 text-generating system and Weinberg realized that it could be used to improve legal workflows.
Harvey, powered by OpenAI’s GPT-4 model family, can answer legal questions phrased in natural language, like “Tell me what the differences are between an employee and independent contractor in the Fourth Circuit” and “Tell me if this clause in a lease is in violation of California law, and if so, rewrite it so it is no longer in violation.” Harvey also offers tools that can automatically extract information from trial transcripts, automatically find legal documents that can back up court arguments and generate first drafts of filings that incorporate info and citations from legal databases.
It’s powerful stuff in theory. But it’s also fraught. Given the sensitive nature of most legal disputes, some lawyers and law firms might be reluctant to give a tool like Harvey access to any case documents. There’s also the matter of language models’ proclivity to spout toxicity and made-up facts, which would be particularly poorly received — if not perjurious — in a court of law.
That’s why Harvey has a disclaimer attached to it: The tool isn’t meant to provide legal advice to nonlawyers and should be used under the supervision of licensed attorneys.
Harvey faces some competition. Casetext uses AI, primarily OpenAI models, to find legal cases and assist with general legal research tasks and brief drafting. More surgical tools like Klarity use AI to strip drudgery from contract review. At one point in time, startup Augrented was even exploring ways to leverage OpenAI models to summarize legal notices or other sources in plain language to help apartment tenants defend their rights.
But Weinberg and Pereyra claim that Harvey is flying high and in use “daily” by tens of thousands of lawyers at law firms and consultancies, including Allen & Overy, Macfarlanes, Ashurst, CMS, Reed Smith and PwC. Annual recurring revenue has tripled since last December, the two co-founders said in the blog post, while Harvey’s workforce has tripled in size.
The Information reported in early June that Harvey hoped to raise $600 million at “at least” a $2 billion valuation, in part to acquire a legal research service called vLex to train its AI products. Those plans fell through — hence the vastly reduced Series C.
Leave a Reply