OpenAI wants to give lawyers an AI research assistant built specifically for the courtroom, not just the open web.
The company on Thursday introduced Astra for Law, a legal-focused configuration of its GPT-6 Astra model for law firms and legal technology companies. It combines the model with a dedicated US legal search index, legal-specific instructions and controls designed for confidential client work.
Early testing shows why specialization matters, but also why human review remains essential. Astra for Law substantially outperformed GPT-6 Astra using web search alone on OpenAI’s legal research benchmark, yet passed the overall correctness check on only 54% of questions.
The research numbers are promising, but limited
OpenAI tested Astra for Law on 200 US legal research questions from Vals AI’s Legal Research Bench private validation set.
At the highest reasoning setting, Astra for Law passed the benchmark’s overall correctness check on 54% of questions, compared with 38.7% for GPT-6 Astra, which used web search alone. OpenAI describes that as a 40% relative improvement. The company also reported that Astra for Law found 24% more reference cases on case-law questions and retrieved up to 54% more relevant passages from the correct court opinions on an audited set.
Those figures, however, come from OpenAI’s own testing and cover a specific validation set. A 54% correctness rate also means the system did not pass the test on nearly half of the questions, underscoring why lawyers still need to verify authorities and conclusions.
OpenAI wants lawyers to keep their existing tools
Astra for Law is not being positioned as a reason for firms to abandon their existing legal software. OpenAI is launching 26 partner-built plugins that connect ChatGPT to tools including Relativity, Clio, iManage, Intapp and DeepJudge. Thomson Reuters is bringing HighQ matter context into ChatGPT and previewing a CoCounsel Legal connector.
Harvey and Legora will also be able to build products using Astra for Law through its API. OpenAI says, “Building on OpenAI should mean getting more from the ecosystem, not replacing it.” That approach matters because legal technology is already fragmented around document management, discovery, research, billing and deal workflows.
The most important part of Astra for Law may be the combination of model, search, firm knowledge and workflow software rather than any single feature.
OpenAI is effectively trying to become the intelligence layer underneath existing legal applications while also giving firms a way to build their own systems. Sullivan & Cromwell, Ropes & Gray, and Cooley have already built custom applications based on their internal methods and expertise.
That could make AI more useful within firms because the technology can operate within established processes rather than forcing lawyers into a generic chatbot. But it also puts OpenAI closer to territory traditionally controlled by legal research and workflow vendors. The company says its index complements licensed products, while firms can continue to use specialist tools via integrations.
More must-read AI coverage
- SS&C Intralinks DealCentre AI vs. Datasite: Which platform is built for the future of dealmaking?
- SS&C Intralinks FundCentre AI vs. Juniper Square: Which platform better supports modern private markets fund managers?
- Why Data, Not Models, Determines AI Success
- The Rise of the AI-Native Factory: How Physical AI Is Transforming Manufacturing
Privacy and availability
OpenAI is initially offering Astra for Law to selected US law firms through its Trusted Access program in ChatGPT and Codex. Eligible firms can receive Zero Data Retention on the API, while ChatGPT Enterprise usage is excluded from human review by default.
OpenAI is also working with Latham & Watkins on information permissions, ethical walls, client instructions and firm oversight.
The API is planned for later release as gpt-6-astra-law, giving legal technology companies another route to build specialized applications. For now, the system remains limited to selected US firms, and its legal search coverage does not extend beyond US law. That geographic limitation will matter for multinational firms as OpenAI expands its legal offering.
What Astra for Law means for law firms
For law firms, the bigger shift is not simply getting another AI research tool. It is deciding how much legal work can be delegated to AI before a lawyer needs to step back into the process.
That could change how firms divide work between associates, partners and technology. Tasks that once required hours of initial research or document analysis may increasingly begin with an AI-generated first pass, leaving attorneys to spend more time testing the reasoning, challenging assumptions and applying professional judgment.
It could also change what firms expect from their technology investments. As legal AI becomes more specialized, firms may increasingly judge these systems by how well they fit into actual matters and reduce repetitive work, rather than by how impressive their answers look in a standalone chatbot.
For legal and IT leaders, workflow design is as important as model capability. The firms that get the most value from tools like Astra for Law may be the ones that determine where AI genuinely saves attorney time, where human review remains mandatory, and how those boundaries should change as the technology improves.
Also read: Google is also pushing deeper into legal AI with Gemini Enterprise for Legal, which brings AI agents to law firms for research, contract review, and other legal workflows.