OpenAI is taking GPT-6 Astra into another highly specialized professional field. The company has launched Astra for Law, a new legal AI foundation built around GPT-6 Astra and designed to help law firms and legal technology companies handle research, analysis, drafting, and other complex legal workflows.

Rather than introducing an entirely new model, OpenAI is configuring GPT-6 Astra with legal-specific tools, settings, context, and instructions. The result is designed to give lawyers a system that can work from client facts, find relevant authorities, develop arguments, examine weaknesses, and assist with consequential legal work while keeping professional judgment with the lawyer.

The launch marks OpenAI’s latest move to build domain-specific versions of its frontier models. After introducing GPT-6 Astra as a general-purpose model for demanding professional work, the company is now adapting that capability around the information sources, workflows, and privacy requirements of particular industries.

A legal search engine built into Astra

One of the biggest differences between Astra for Law and a general-purpose AI assistant is its dedicated Legal Search Index.

OpenAI says the index can search U.S. case law, statutes, regulations, court rules, and administrative decisions across more than 230 million URLs, with new sources added daily. The company has also worked with the Free Law Project to bring CourtListener’s case-law collection into the research experience. OpenAI says that collection covers more than 99.9% of published U.S. precedential case law.

That matters because legal research isn’t simply about finding information that looks relevant. Lawyers need to identify the right authority, locate the passages that actually support a proposition, understand how an authority applies to a particular set of facts, and account for cases that may undermine an argument.

OpenAI says Astra for Law is designed around that process. Its legal instructions can help distinguish a court’s holding from other observations, identify cases that weaken an argument, and explain how contractual provisions can shift risk between parties.

The system is therefore intended to move beyond the familiar “ask a chatbot a legal question” workflow. Instead, OpenAI is positioning Astra for Law as an underlying foundation that legal professionals and technology companies can build into more specialized applications.

OpenAI reports stronger legal research performance

OpenAI also tested whether the specialized configuration actually improves legal research.

According to the company’s evaluation, Astra for Law was tested 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% of questions, compared with 38.7% for GPT-6 Astra using web search alone.

OpenAI says the legal configuration also found 24% more reference cases on case-law-focused questions. On an audited set of target passages, it retrieved up to 54% more relevant passages from the correct court opinions compared with GPT-6 Astra using web search alone.

These figures come from OpenAI’s own evaluation and should be understood in that context. They demonstrate the improvement OpenAI observed on its selected test set rather than establishing that Astra for Law will perform at the same level across every type of legal matter.

Still, the direction is significant. OpenAI is effectively arguing that giving a general reasoning model a purpose-built legal information layer can make a meaningful difference in the kind of research lawyers actually need to perform.

Built for legal analysis, not just legal search

Research is only one part of legal work. Astra for Law also includes custom instructions for legal analysis and writing.

The system can take information gathered during research and apply it to a client’s facts, help develop arguments or deal terms, identify weaknesses, and consider uncertainty. That could make it useful for everything from assessing a dispute to preparing an initial analysis of a transaction.

The distinction is important. OpenAI isn’t presenting the technology as an autonomous replacement for a lawyer. Its own description emphasizes supporting the expertise and judgment of lawyers and legal technology companies.

That human-review model also reflects the stakes involved. A legal AI system can help surface authorities, organize large amounts of information, and produce a useful first draft, but the consequences of relying on an incorrect citation, misunderstood precedent, or flawed interpretation can be substantial.

OpenAI’s Help Center explicitly tells users to review Astra for Law’s answers and cited sources before relying on them.

OpenAI wants legal technology companies to build on Astra

Astra for Law isn’t only aimed at lawyers using ChatGPT directly.

OpenAI says API customers, including Harvey and Legora, will be able to build products and workflows on top of the legal foundation. That could allow legal technology companies to incorporate GPT-6 Astra’s reasoning capabilities without having to recreate the underlying legal search and instruction layer themselves.

OpenAI is also expanding integrations with software already used by law firms. The company says it has introduced 26 ecosystem plugins connecting ChatGPT with specialist tools, including platforms such as Relativity and Clio. Other integrations involve providers including iManage, Intapp, DeepJudge, and Thomson Reuters.

That approach could be important for adoption. Large law firms already have established document-management systems, research databases, matter-management tools, and internal knowledge repositories. An AI system that can fit into those environments has a very different proposition from a standalone chatbot that requires lawyers to move their work somewhere else.

Privacy and controlled access are a major part of the launch

Legal work also comes with unusually strict confidentiality requirements, and OpenAI is addressing that as part of Astra for Law.

The company says it is expanding privacy and governance controls for confidential client work. Astra for Law will initially be available to selected law firms through OpenAI’s Trusted Access program in ChatGPT and Codex, with API availability coming soon. The model appears in the model picker as GPT-6 Astra Law.

OpenAI has been working with firms including Sullivan & Cromwell, Ropes & Gray, Cooley, Latham & Watkins, and Wachtell Lipton as it develops legal AI applications.

For legal organizations, the bigger story may therefore be less about having another AI chatbot and more about getting an AI foundation that can operate within existing professional workflows and governance requirements.

Another step toward industry-specific AI

Astra for Law follows a broader strategy from OpenAI of adapting its frontier models for professional industries rather than treating one general-purpose model as the answer to every workflow.

GPT-6 Astra already includes capabilities aimed at complex professional work, including research, computer use, document creation, and multistep workflows. Astra for Law adds the domain-specific information and instructions needed to make those capabilities more useful in a legal environment.

The timing also puts OpenAI into a rapidly developing legal AI market. Other major AI companies and legal technology providers are pursuing their own combinations of foundation models, proprietary legal data, research tools, and workflow integrations. Reuters reported that OpenAI’s launch is part of a broader push by major AI companies to win law-firm and legal-department customers.

For now, Astra for Law is being introduced as a specialized foundation rather than a broadly available consumer product. Selected U.S. law firms can access it through Trusted Access, while API availability is expected later.

The bigger idea is straightforward: OpenAI wants GPT-6 Astra to become more than a general AI model. With Astra for Law, it is adding the legal sources, instructions, integrations, and controls needed to turn that general intelligence into a foundation for professional legal work.