A legal foundation built around GPT-6 Astra

OpenAI says Astra for Law is a new foundation for law firms and legal technology companies to build AI products and workflows around their own expertise. It combines GPT-6 Astra, described by the company as its latest and most powerful model, with settings, tools and context tailored to professional legal work.

The system is intended to support more than general-purpose question answering. OpenAI has paired the model with instructions for legal analysis and writing, allowing firms and technology providers to adapt the underlying intelligence to their own applications, methods and workflows. API customers including Harvey and Legora will be able to build on Astra for Law, while OpenAI says it plans to bring these legal capabilities to future versions of its latest models.

Legal search across a large U.S. corpus

Astra for Law includes a legal search index designed to help with the early stages of research: finding the right authority, locating relevant passages and assessing how applicable and binding those authorities may be. OpenAI says the index complements licensed content and specialist products that firms use from providers such as Thomson Reuters.

According to OpenAI, the index can search U.S. case law, statutes, regulations, court rules and administrative decisions across a corpus of more than 230 million URLs, with sources added daily. The company also worked with Free Law Project, the nonprofit behind CourtListener, to bring a case-law collection covering more than 99.9% of published U.S. precedential case law into the research experience.

OpenAI tested the complete Astra for Law configuration on 200 U.S. legal research questions from the private validation set of Vals AI’s Legal Research Bench. At the highest reasoning effort for both systems, 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. OpenAI characterizes that result as a 40% relative improvement.

OpenAI reports broader research results

The company also says Astra for Law produced more comprehensive answers in the evaluation. On case-law-focused questions, it found 24% more reference cases than GPT-6 Astra using web search alone at the highest reasoning effort. On an audited set of target passages, it retrieved up to 54% more relevant passages from the correct court opinions when the systems were compared at the same reasoning effort.

OpenAI presents the results as evidence of a stronger research foundation for tasks such as advising on a deal, assessing a dispute or developing a legal strategy. The company says the system is designed to provide authorities that lawyers can examine themselves, rather than treating the model’s answer as a substitute for professional review.

From research to firm-specific workflows

Legal-specific instructions are intended to guide Astra for Law beyond source retrieval. OpenAI says the system can apply research to a client’s facts, develop arguments or deal terms, and identify weaknesses and uncertainty. This includes distinguishing a court’s holding from other observations, addressing cases that weaken an argument and explaining how a contract exception changes the allocation of risk between parties.

Selected firms have worked with OpenAI’s forward-deployed engineers to adapt ChatGPT Enterprise with custom interfaces and integrations for proprietary data. Sullivan & Cromwell built an agreement analyzer that uses its negotiating playbooks and selected precedents to identify risks arising when provisions are read together, then proposes redlines and draft client advice for lawyers to review and refine.

Ropes & Gray built a deal diligence system around its lawyers’ approach to reviewing a data room and determining what matters to a transaction. The system can trace findings to their source and identify questions that may affect an acquisition, including whether important customer contracts require notice or consent. Cooley built GO Public to apply its capital-markets expertise to IPO preparation, from drafting the filing to identifying risks for management’s attention and carrying changes through the filing for combined review.

Confidentiality controls and availability

OpenAI says it is expanding privacy and governance measures to give firms specific controls for confidential client work. Astra for Law can be customized with 26 new ecosystem plugins that connect ChatGPT to specialist tools already used by firms, including Relativity and Clio.

The company is working with Latham & Watkins on information permissions, ethical walls, client instructions and firm oversight. Astra for Law will initially be offered to selected law firms through a Trusted Access program in ChatGPT and Codex. Eligible firms will receive access for lawyers and people working under their supervision for professional legal work, with Zero Data Retention on the API and ChatGPT Enterprise excluded from human review by default.

OpenAI says the model will appear in ChatGPT’s model picker as GPT-6 Astra Law and will be available in the API as gpt-6-astra-law when the API rollout begins. The company describes the broader goal as enabling firms to turn their precedents, methods and professional judgment into AI tools and workflows built to their own standards.

Sources

• OpenAI NewsIntroducing Astra for Law