Skip to content
NewsProduct launch

OpenAI Launches Astra for Law, Adding a 230-Million-Source US Legal Index to GPT-6 Astra

· by Pondero Newsdesk

The short version

OpenAI introduced Astra for Law on September 17, 2026, a GPT-6 Astra configuration backed by a 230-million-URL US legal index built on CourtListener and updated daily, benchmarking 54% accuracy versus 38.7% for base GPT-6 Astra on external legal research questions.

OpenAI Launches Astra for Law, Adding a 230-Million-Source US Legal Index to GPT-6 Astra

Law firms using ChatGPT now have direct access to a proprietary legal research index: OpenAI launched Astra for Law on September 17, 2026, adding a 230-million-URL database of US case law and regulations to GPT-6 Astra and recording a 40% benchmark improvement over using the same model with web search alone, per Artificial Lawyer. Latham & Watkins, Ropes & Gray, Cooley, and Sullivan & Cromwell joined as early adopters through the Trusted Access program.

What

Astra for Law pairs GPT-6 Astra with legal-specific instructions and OpenAI's new Legal Search Index. The index covers US case law, statutes, regulations, court rules, and administrative decisions across more than 230 million URLs, with sources added daily. The case-law foundation draws from CourtListener, the database maintained by the Free Law Project, covering more than 99.9% of published US precedential case law, per Artificial Lawyer.

OpenAI measured the configuration against GPT-6 Astra with web search on 200 US legal research questions from Vals AI's Legal Research Bench. Astra for Law passed the overall correctness check on 54.0% of questions; the base model with web search passed 38.7%, a 40% relative improvement. On case-law-focused questions, Astra for Law found 24% more reference cases and retrieved up to 54% more relevant passages from the correct court opinions at the same reasoning effort, per the Artificial Lawyer report.

Initial access runs through the Trusted Access program inside ChatGPT and Codex, with API access listed as coming soon. For eligible firms, the offering includes Zero Data Retention on the API and, for ChatGPT Enterprise users, exclusion from human review by default. Pricing was not disclosed.

The launch includes 26 partner-built plugins and 47 community plugins. Named integration partners are Thomson Reuters, Harvey, Legora, and iManage. Latham & Watkins collaborated with OpenAI specifically on information permissions, ethical walls, and firm oversight design.

Why it matters

The meaningful change here is structural, not incremental. Thomson Reuters (Westlaw) and LexisNexis have long held the default position in law-firm research workflows partly because their proprietary databases were not available through general-purpose AI models. Astra for Law puts a competing index behind a model many firms already deploy inside ChatGPT Enterprise. The integration cost for a firm already on that plan is lower than adopting a separate legal AI platform with its own contract, data pipeline, and training curve.

The 54.0% benchmark result deserves scrutiny before firms draw conclusions. Vals AI's Legal Research Bench is a third-party evaluation, which is the right kind of citation, but a private validation set is harder to replicate independently. The 40% relative improvement over base GPT-6 is still the most concrete signal released: it quantifies the gap between a configured legal product and a general model, and firms can use that gap to decide whether internal pilots are warranted.

For legal AI buyers: the first practical question is whether Astra for Law's output on routine matter research reduces billable hours spent on Westlaw or Lexis queries at all. OpenAI has not provided workflow integration data or per-matter pricing. That means internal pilots are necessary before any cost comparison is real.

What to watch next

Three disclosures will determine how broadly Astra for Law can be adopted. Pricing for the Trusted Access program has not been released; what OpenAI charges firms relative to existing Westlaw and Lexis subscriptions sets the commercial stakes. API access, listed as coming soon, matters most to legal-tech builders who want to integrate the Legal Search Index into their own products. And competitive responses from Thomson Reuters and LexisNexis, both of which have AI products running on overlapping timelines, will establish whether the 40% benchmark advantage translates into a durable market position.

Sources