OpenAI on September 17, 2026 launched Astra for Law, a legal-focused configuration of its GPT-6 Astra model bundled with a dedicated legal search index, Trusted Access controls and a plugin ecosystem targeting large law firms and legal-tech vendors. The launch escalates OpenAI's push into vertical AI and puts it in direct competition with vertical incumbents such as Harvey and Thomson Reuters' Ask Practical Law.
A Purpose-Built Legal Search
Astra for Law ships with a legal search index covering 230+ million URLs across U.S. case law, statutes, regulations and administrative decisions, sourced in part through a collaboration with the Free Law Project and its CourtListener corpus. OpenAI says the index covers 99.9% of published U.S. precedential case law and lifts the model to 54% correctness on Legal Research Bench, up from 38.7% for GPT-6 with web search only. The new tooling retrieves 24% more relevant case law references and up to 54% more relevant passages than its own baseline.
Trusted Access, Governance And ZDR
Astra for Law is initially available through OpenAI's Trusted Access program for selected firms, with API access "coming soon." The stack includes Zero Data Retention on the API for eligible firms and excludes ChatGPT Enterprise usage from human review by default. OpenAI is co-developing a governance framework with Latham & Watkins, and named custom-tool collaborators include Sullivan & Cromwell, Ropes & Gray, Cooley, Skadden and Wachtell, Lipton, Rosen & Katz.
Harvey And Legora Get API Access
Two of the most closely watched legal-AI companies, Harvey and Legora, are the launch API customers, joining a partner-plugin ecosystem that already spans Relativity, Clio, iManage and Thomson Reuters HighQ. Twenty-six partner-built plugins and nine community plugins are live at launch, plus 47 adaptable skills.
Why It Matters
OpenAI is doubling down on the vertical-AI thesis after the enterprise agentic land grab we detailed in Salesforce's Koa launch and Anthropic's R&D Automation Index. For litigators, the pitch is a single stack that combines a frontier reasoning model with authoritative legal search and firm-level controls; for OpenAI, it is a case study in vertical customisation that puts an existing legal-AI ecosystem on notice.
Reporting based on coverage from OpenAI, SiliconANGLE, Unite.AI, Artificial Lawyer and Analytics Insight.