OpenAI launches Astra for Law with a 230M-URL legal search index and Trusted Access

OpenAI launched Astra for Law on September 17, pairing its GPT-6 Astra model with a legal search index covering over 230 million URLs of US law. The system is offered first to selected law firms through a Trusted Access program in ChatGPT and Codex, with an API rollout planned later. For AI builders, the launch is less about the model itself and more about how OpenAI is packaging governance, data controls, and ecosystem integrations to sell into a risk-averse professional market.

What OpenAI actually shipped

Astra for Law is GPT-6 Astra configured with legal search, instructions for legal analysis and writing, and governance controls aimed at law firms. The underlying legal search index draws heavily from the Free Law Project's CourtListener collection, which OpenAI says covers more than 99.9% of published US precedential case law. The index also includes statutes, regulations, court rules, and administrative decisions, with sources added daily.

OpenAI tested the configuration on 200 US legal research questions from Vals AI's Legal Research Bench. At the highest reasoning effort, Astra for Law passed the overall correctness check on 54.0% of questions, compared to 38.7% for GPT-6 Astra with web search alone. OpenAI presents that as a 40% relative improvement. It also means the system fails the correctness check on nearly half the questions. The testing is OpenAI's own and has not been independently audited.

Alongside the model, OpenAI launched 26 partner-built plugins connecting ChatGPT to tools firms already use, including Relativity, Clio, iManage, Everlaw, and others. Thomson Reuters is bringing HighQ matter context into ChatGPT and previewing a CoCounsel Legal connector. Harvey and Legora, two well-funded legal AI startups, will build on Astra for Law through the API.

Why the governance wrapper matters for builders

The most interesting part for AI builders is the governance infrastructure. Eligible Trusted Access users get Zero Data Retention on the API. OpenAI is working with Latham & Watkins on information permissions, ethical walls, and client instructions. Forward-deployed engineers are inside firms like Sullivan and Cromwell, Ropes and Gray, and Cooley, building custom tools on top of the platform.

This is the same template OpenAI used a week earlier for financial services: a frontier model, licensed or indexed data, a governance wrapper, and partner connectors. For builders shipping AI into regulated industries, the pattern is worth studying. The value is not just the model but the ability to deploy it inside existing compliance and oversight frameworks.

What changes for legal tech developers

If you are building legal AI products, the launch changes the competitive landscape in two ways. First, OpenAI now owns the model, the research index, the distribution surface (ChatGPT and Codex), and the plugin directory. Specialists like Harvey and Legora will build on OpenAI's infrastructure, but they also compete with OpenAI's direct-to-firm deployments. Second, the 26 partner plugins show a clear integration pattern: vendors keep their substantive analysis inside their own platforms while letting ChatGPT handle the query surface. Time entry stays in Laurel, deeper workflows stay in Litera, and Relativity keeps its analysis inside its own platform. Each vendor is describing where the subscription stays.

What OpenAI hasn't said yet

Several important details are missing. Pricing was not disclosed. The list of firms in Trusted Access is not public. The index covers US law only, and OpenAI has not addressed European governance requirements or rollout timing. The benchmark results are self-reported and should be treated as vendor claims until independently verified. For builders evaluating the API, the lack of pricing and regional availability makes it hard to plan around.