OpenAI has entered the legal industry with a dedicated tool built for lawyers, and the move signals just how serious the company is about turning artificial intelligence into a fixture of everyday law firm work. The new product, called Astra for Law, marks OpenAI’s clearest bid yet to become the default OpenAI legal AI provider for America’s largest firms, pairing a specialized model with a legal search index designed to cut down on the kind of made-up citations that have embarrassed lawyers using generic chatbots.
Key takeaways
- OpenAI launched Astra for Law, a version of GPT-6 Astra tailored specifically for legal work, according to Business Insider and the-decoder.
- The product includes a legal search index covering US case law, statutes, and regulations across more than 230 million URLs, built on data from the Free Law Project that covers over 99.9 percent of published US precedents.
- In OpenAI’s own test using Vals AI’s Legal Research Bench, Astra for Law passed 54 percent of 200 questions, compared with 38.7 percent for GPT-6 Astra using plain web search.
- Law firms get a Trusted Access program with privacy controls including zero data retention, while API customers like Harvey and Legora can build on top of the model.
- OpenAI is targeting the AmLaw 200, the largest American law firms, and is launching 26 plugins for tools such as Relativity, Clio, Thomson Reuters, iManage, and DeepJudge.
OpenAI launches Astra for Law targeting US legal market
Astra for Law is a version of OpenAI’s most advanced model, GPT-6 Astra, adapted specifically for legal analysis and writing. Rather than releasing a generic assistant, OpenAI paired the model with instructions tuned for legal reasoning and a purpose-built search index, according to reporting from the-decoder.
Jason Boehmig, who leads legal product at OpenAI, framed the strategy around customer focus rather than competitive positioning. “You win by focusing on the customer and not the competition,” he said at a press briefing, as reported by Business Insider. He added that the challenge was balancing raw model capability with relevance to a specific profession: “What we really solved is: How do you stay on the frontier of capability, but also make the output relevant to a specific type of knowledge work.”
Sherwin Wu, a member of OpenAI’s technical staff, said the adjustments ground the model in “actual case law,” which he said led to fewer hallucinations and “just much more sound legal reasoning.” Business Insider also reported that Astra for Law’s answers run roughly twice as long as typical Astra responses. Boehmig said that length reflects direct user feedback: “Lawyers like longer responses. That was one thing that came back from feedback.”
Extensive legal data and benchmark performance
The strength of Astra for Law rests largely on the depth of its legal index and how that translates into measurable accuracy gains over a general-purpose model. The search index spans US case law, statutes, and regulations across more than 230 million URLs, drawing on data supplied by the Free Law Project, a source that claims to cover over 99.9 percent of published US legal precedents.
To test whether that specialized data actually improves legal research, OpenAI ran its own evaluation using Vals AI’s Legal Research Bench. Astra for Law passed 54 percent of 200 test questions. GPT-6 Astra using only plain web search, by contrast, passed just 38.7 percent of the same questions. That gap of roughly 15 percentage points is the clearest evidence OpenAI has offered so far that a domain-specific search index, rather than the base model alone, drives meaningful improvement on legal tasks.
It’s worth noting that the benchmark figures come from a test OpenAI conducted itself, which means outside verification of the results is not yet available. Still, the scale of the gap suggests the legal-specific tooling is doing real work rather than serving as a marketing veneer.
Privacy controls and integration with legal technology ecosystem
For law firms, the appeal of a legal AI tool depends as much on confidentiality as accuracy, and OpenAI is trying to address that directly. Firms adopting Astra for Law get access to a Trusted Access program that includes privacy safeguards such as zero data retention, an important consideration for an industry bound by strict confidentiality obligations.
OpenAI is also opening the door to developers. API customers such as Harvey and Legora can build their own applications directly on top of Astra for Law, extending the model beyond OpenAI’s own interface. Business Insider reported that the launch also comes bundled with ChatGPT plugins from Harvey and Legora themselves, alongside integrations from Thomson Reuters, iManage, and DeepJudge.
Beyond those partnerships, OpenAI is rolling out 26 plugins compatible with widely used legal software including Relativity and Clio. This why-it-matters angle is straightforward: by embedding itself inside the tools law firms already use daily, OpenAI reduces the friction that often stalls enterprise AI adoption, instead of asking firms to overhaul their existing workflows.
Sherwin Wu also indicated that a citations feature developed for OpenAI’s recent financial services product will soon be brought over to the legal offering, a sign that the company is cross-pollinating tools built for different regulated industries.
Competitive landscape and strategic implications
OpenAI is explicitly targeting the AmLaw 200, the roster of the 200 largest American law firms, according to Boehmig. That focus places Astra for Law squarely in competition with Anthropic, which expanded its own legal AI push in May with a release built around similar integrations and legal-specific tooling.
The timing also lines up with a broader pattern. Business Insider notes that OpenAI has been rolling out industry-specific products across several sectors recently, including a financial services tool aimed at Wall Street, along with offerings tailored for teachers, clinicians, and life sciences researchers. That sequence suggests a company preparing its commercial footprint across regulated, high-value professional markets rather than treating legal work as an isolated experiment.
For law firms and legal tech vendors, the practical question now is how quickly Astra for Law’s benchmark advantage translates into real deployments, and whether Anthropic’s competing push in the same market narrows that gap. Either way, the entrance of two major AI labs into legal research tools signals that this corner of professional services is becoming one of the more contested fronts in the broader AI competition.
FAQ
What is Astra for Law?
Astra for Law is OpenAI’s legal AI product based on GPT-6 Astra, designed specifically for US legal work with a specialized legal search index.
What data does Astra for Law use for legal research?
It uses data from the Free Law Project, covering over 99.9% of published US precedents, including case law, statutes, and regulations.
How does Astra for Law perform compared to GPT-6 Astra with plain web search?
Astra for Law passed 54% of questions in Vals AI’s Legal Research Bench test, outperforming GPT-6 Astra with plain web search, which passed 38.7%.
How does OpenAI ensure privacy for law firms using Astra for Law?
Law firms receive a Trusted Access program that includes privacy controls such as zero data retention.
Article produced with the assistance of artificial intelligence and reviewed by the editorial team.