This week, OpenAI introduced Astra for Law, a version of its most powerful model configured specifically for professional legal work. Astra for Law pairs GPT-6 Astra with a new legal search index, custom instructions for legal analysis and writing, and governance controls built for confidential client matters. Law firms and legal technology companies can build their own products and workflows on top of it.
What Astra for Law is
Astra for Law is not a single app. OpenAI describes it as a foundation. GPT-6 Astra is combined with settings, tools and context tailored to legal practice. The configuration includes a legal search index for finding authorities, instructions that steer the model toward careful legal analysis and writing, and controls that let firms decide how AI is used in their matters.
Access starts through a Trusted Access Program for selected firms, inside ChatGPT and Codex. In the model picker it appears as GPT-6 Astra Law, and in the API it will carry the identifier gpt-6-astra-law. API access is coming soon, and legal technology companies including Harvey and Legora will build on it, bringing the model’s capabilities into their own products. Alongside the launch, 26 new ecosystem plugins connect ChatGPT to specialist tools firms already use, such as Relativity and Clio.
A legal search index built for real research
The legal search index is the centerpiece. Legal research starts with finding the right authority, locating the relevant passages and understanding how binding they are for the matter at hand. The index lets Astra for Law search U.S. case law, statutes, regulations, court rules and administrative decisions across a corpus of more than 230 million URLs, with new sources added daily. A collaboration with Free Law Project, the nonprofit behind CourtListener, brings a case-law collection covering more than 99.9% of published U.S. precedential case law into the research experience.
OpenAI positions the index as a complement to the licensed content and specialist products firms already rely on from providers such as Thomson Reuters, not a replacement.
To measure the difference, OpenAI tested the full Astra for Law setup on 200 U.S. legal research questions from the private validation set of Vals AI’s Legal Research Bench. At the highest reasoning effort, Astra for Law passed the benchmark’s overall correctness check on 54.0% of questions, compared with 38.7% for GPT-6 Astra using web search alone. That is a 40% relative improvement, and the answers were more comprehensive. On case-law-focused questions, the system found 24% more reference cases, and it retrieved up to 54% more relevant passages from the correct court opinions at equal reasoning effort.
The practical result is a research foundation you can verify. Answers arrive with authorities and passages attached, so you can examine the sources yourself instead of taking the model’s word for it.
From research to end-to-end legal work
Research is only the first step in a matter. The custom instructions in Astra for Law guide the model through applying that research to client facts, developing arguments or deal terms, and flagging weaknesses and uncertainty. In practice that means distinguishing a court’s holding from its other observations, addressing cases that cut against an argument, or explaining how a contract exception shifts risk between parties.
When prompted to identify good law with similar fact patterns, Astra for Law pinpointed relevant precedent and matched fact patterns better than other frontier models, according to OpenAI’s testing. Early users point in the same direction. Niko Grupen, Head of Applied Research at Harvey, said the system showed strength in “grounding answers in on-point authorities, citing with precision, and offering practical, advisory guidance.”
Legal-grade trust and governance
Confidentiality decides whether a firm can use AI at all. For eligible firms, the Trusted Access Program includes Zero Data Retention on the API, and usage of ChatGPT Enterprise is excluded from human review by default. Lawyers and people working under their supervision get access for professional legal work under firm-controlled conditions.
OpenAI is also working with Latham & Watkins on the governance layer, covering information permissions, ethical walls, client instructions and firm oversight. Michael Rubin, Chair of Latham’s AI Strategy Committee, framed the collaboration as part of the firm’s “enterprise-wide strategy for responsible AI,” adding that more capable AI raises the bar for “rigorous governance, oversight, and accountability.”
What firms are already building with it
The most telling part of the launch is what early adopters have built. OpenAI’s forward-deployed engineers worked with selected firms to adapt ChatGPT Enterprise with custom interfaces and integrations to proprietary data.
- Sullivan & Cromwell built an agreement analyzer that brings the firm’s negotiating playbooks and selected precedents into the review of a new deal. It spots risks that emerge when provisions are read together, then turns those findings into proposed redlines and draft client advice lawyers can challenge and refine.
- Ropes & Gray built a deal diligence system around how its lawyers work through a data room. It traces findings back to the source and flags questions that could affect an acquisition, such as whether key customer contracts require notice or consent.
- Cooley built GO Public, which carries its capital markets expertise into IPO preparation, from drafting the filing to identifying risks that deserve management’s attention. When a deal changes, the change propagates across the filing so lawyers review the implications together.
- Skadden is designing tools to help clients assess regulatory risk as they pursue transactions or bring products to market.
- Wachtell, Lipton, Rosen & Katz is exploring how the technology can support litigation and corporate judgment.
Dave Peinsipp, partner and co-chair of Cooley’s global capital markets group, said the collaboration lets his firm move lawyers “more quickly through intensive preparation and into questions that require judgment, market experience and strategic thinking.”
An open ecosystem of legal plugins
The 26 partner-built plugins cover both the practice and the business of law. With iManage, a lawyer can draft a negotiation brief in ChatGPT and save it to the matter file. Intapp can surface activities that may need a time entry for review. DeepJudge can pull prior deals into a comparison. Thomson Reuters is bringing HighQ matter context into ChatGPT and previewing a forthcoming CoCounsel Legal connector. Its CTO Joel Hron said legal professionals need “trusted intelligence, relevant enterprise and matter context, purpose built legal capabilities, and the governance required for high stakes work.”
The launch also includes 9 community plugins from lawyers and legal engineers at LegalQuants, LECG and Skills.law, containing 47 custom skills that practitioners can adapt or extend. And ChatGPT for Word is now generally available, so lawyers can proofread, receive suggested edits and flag formatting issues in the tool where most drafting already happens.
The philosophy is deliberately open and composable. Firms keep their specialist products, bring their own knowledge, and assemble the pieces that fit their practice.
What Astra for Law means for your firm
Two caveats deserve attention. First, the search index is U.S.-focused, so firms working primarily in other jurisdictions will find the research layer less complete for now. Second, the benchmark results cut both ways. A 54% correctness score is a large improvement, yet it also means close to half of the research answers in the evaluation failed the overall correctness check. Verification stays part of the job.
There is also a strategic tension worth watching. Harvey and Legora will build on Astra for Law while OpenAI moves closer to the legal work itself, which makes the company both platform provider and potential competitor to its own ecosystem.
The firms that gain the most will be those that treat Astra for Law as infrastructure for their own expertise, as Sullivan & Cromwell, Ropes & Gray and Cooley have done, rather than as a finished product. The model supplies the research depth and the reasoning. The judgment still comes from you.
