OpenAI Launches Astra for Law: A 230M-URL Legal Search Index Scoring 54.0% on a Legal Research Benchmark
Announced September 17, Astra for Law bundles GPT-6 Astra with a legal search index and legal drafting instructions; on Vals AI's Legal Research Bench it passes overall correctness at 54.0% versus 38.7% for web search alone.
In this story
On September 17, 2026 OpenAI introduced Astra for Law, framed as "a new foundation for law firms and legal technology companies": GPT-6 Astra combined with a legal search index and instructions tailored to legal analysis and writing. It appears in the model picker as "GPT-6 Astra Law" and in the API as gpt-6-astra-law, initially offered to selected firms through Trusted Access with the API version coming soon. API customers including Harvey and Legora are named as builders on top of it.
What the search index covers
OpenAI says Astra for Law can search U.S. case law, statutes, regulations, court rules and administrative decisions across a corpus of more than 230 million URLs with sources added daily, and that its work with Free Law Project — the nonprofit behind CourtListener — brings that organization's case-law collection, covering more than 99.9% of published U.S. precedential case law, into the research experience. The post presents the index as complementing licensed content and specialist products firms buy from providers such as Thomson Reuters.
The gap OpenAI measured
Using 200 U.S. legal research questions from the private validation set of Vals AI's Legal Research Bench, with both systems at the highest reasoning effort:
- Astra for Law passed the evaluation's overall correctness check on 54.0% of questions, versus 38.7% for GPT-6 Astra using web search alone — which OpenAI describes as a 40% relative improvement.
- On case-law-focused questions it found 24% more reference cases than GPT-6 Astra with web search alone.
- On the audited set of target passages it retrieved up to 54% more relevant passages from the correct court opinions at the same reasoning effort.
Privacy, governance and firm-built tools
A dedicated Trusted Access Program for eligible firms includes Zero Data Retention on the API, with ChatGPT Enterprise usage excluded from human review by default. OpenAI says it is working with Latham & Watkins to design for information permissions, ethical walls, client instructions and firm oversight. The same announcement lists 26 partner-built plugins spanning iManage, Intapp, DeepJudge and Thomson Reuters (HighQ matter context now, with a CoCounsel Legal connector previewed), plus 9 community plugins from LegalQuants, LECG and Skills.law with 47 reusable custom skills; ChatGPT for Word became generally available the same day.
Firm-built examples include an agreement analyzer from Sullivan & Cromwell and a deal diligence system from Ropes & Gray. A Cooley customer story published the same day describes its GO Public offering, built on ChatGPT Work's agentic harness to automate the first pass of IPO preparation, and notes Cooley advised on 180 deals globally in 2025 totaling more than $51.5 billion in deal volume.
This article aggregates official announcements and public reporting; original sources are linked below.
Source:OpenAI
AI Tools Daily is a bilingual newsroom covering AI tool launches, product updates and industry trends. Editorial standards · Report a correction
