On September 17, 2026, OpenAI introduced Astra for Law, a specialized version of its GPT-6 model configured for legal analysis and research, supported by a massive new US legal search index developed in partnership with the Free Law Project and others.
- Astra for Law uses GPT-6 with a US legal search index of 230M+ URLs
- OpenAI partners with legal tech firms to integrate AI into workflows
- Independent audits pending as Astra improves but still imperfect
What happened
OpenAI launched Astra for Law on September 17, a customized GPT-6 model equipped with legal-specific functionality for searching, analyzing, and drafting legal documents. This model is initially available to select firms through a Trusted Access program on ChatGPT and Codex, with an API release planned later. Underpinning Astra for Law is a newly built legal research index covering United States case law, statutes, regulations, court rules, and administrative decisions. This index includes over 230 million URLs and daily updates, with large contributions from the Free Law Project's CourtListener database, claimed to cover more than 99.9% of published US precedential case law.
Alongside the model release, OpenAI introduced 26 partner-built plugins designed to connect ChatGPT to established legal tools such as Relativity, Clio, iManage, and Intapp. Legal vendors including Thomson Reuters, iManage, and Everlaw immediately announced integrations and enhancements to work with Astra, signaling broad industry collaboration. OpenAI highlighted Astra’s improvements in internal testing, reporting a 40% relative boost in legal reasoning correctness compared to GPT-6 Astra with web search alone, and an increased ability to find relevant case references and passages.
Why it matters
The legal profession relies heavily on specialized, licensed legal content and workflow tools. OpenAI’s Astra for Law seeks to complement, not replace, existing legal research platforms like Thomson Reuters’ Westlaw and CoCounsel by providing a powerful AI-driven interface coupled with an expansive free legal data index. This initiative potentially accelerates law firms’ adoption of AI by addressing the need for both comprehensive data access and tailored AI instructions designed for legal professionals.
The launch also marks a key step in the evolving legal tech landscape where AI models are increasingly integrated with traditional software ecosystems. By enabling interoperability through plugins and APIs, OpenAI's approach aims to preserve firms’ existing data governance and subscription boundaries, while enhancing substantive legal research capabilities. However, Astra’s reported error rates indicate that AI is not yet fully reliable for critical legal decision-making, underscoring the ongoing importance of human oversight.
What to watch next
Future developments will focus on expanding Astra for Law’s availability beyond the early trusted access phase and observing how well the legal industry adopts and integrates the model into existing workflows. OpenAI’s claims on performance improvements should be independently audited to verify the model’s reliability and accuracy in practical settings. The results of such evaluations will be crucial for gaining broader confidence among legal professionals.
Another key element to monitor is how third-party legal AI startups and vendors like Harvey and Legora leverage the Astra for Law API to build specialized applications and whether this approach leads to an ecosystem that truly enhances productivity without fragmenting legal workflows. Finally, continued integration announcements and plugin innovations from legal technology providers will signal how well OpenAI’s strategy promotes sustainable collaboration within the legal market.