OpenAI is packaging GPT-6 Astra as a vertical legal product with its own U.S. case-law index, after a similar financial-services launch, according to multiple trade outlets covering a Sept. 17, 2026 announcement. OpenAI’s own launch pages and help article were not independently retrievable for this story.
The offering includes a Legal Search Index covering U.S. case law, statutes, regulations, court rules, and administrative decisions across more than 230 million URLs, LawNext, The Next Web, and Artificial Lawyer reported. The Next Web and Artificial Lawyer said sources are added daily. Coverage describes United States law only.
Access is initially through a Trusted Access program in ChatGPT and Codex for selected law firms. LawNext said the program targets Am Law 200 firms. SiliconANGLE reported that the configuration appears in the model picker as GPT-6 Astra Law and that an API version, gpt-6-astra-law, is due later, with no date or pricing attached. LawNext said the API would be available soon to customers including Harvey and Legora. The Next Web reported that OpenAI has not named all firms inside Trusted Access or said when the product reaches Europe.
Alongside the launch, OpenAI introduced 26 partner-built plugins for legal work, including from Thomson Reuters, Harvey, Legora, and iManage, according to LawNext, SiliconANGLE, and Artificial Lawyer. SiliconANGLE said the plugins also connect ChatGPT to tools such as Relativity, Clio, and DeepJudge, and that Thomson Reuters is bringing HighQ matter context into ChatGPT and previewing a connector for CoCounsel Legal.
Legaltech News reported co-development details with Sullivan & Cromwell, Ropes & Gray, Skadden, and Cooley. LawNext and SiliconANGLE described firm-built workflows such as an agreement analyzer at Sullivan & Cromwell, an M&A diligence system at Ropes & Gray, and Cooley’s GO Public IPO tooling. Firm roles differ slightly across reports.
On Vals legal research questions, OpenAI reported that Astra for Law passed the overall correctness check on 54.0% of questions at the highest reasoning effort, compared with 38.7% for GPT-6 Astra using web search alone — a 15.3-point absolute gain, or about 40% relative improvement, LawNext and The Next Web reported. The Next Web noted the configured system still fails the correctness check on nearly half the questions. Those figures are OpenAI’s own private comparison and have not been independently audited.
If the reporting holds, OpenAI is selling a legal-research configuration of its frontier model to large firms while also supplying legal AI vendors those firms already buy from. The business story is distribution: plugins and Trusted Access pull work into ChatGPT, the API lets specialists keep their products, and a vendor-run 54% Vals score still leaves lawyer review and governance as the buying constraint.