OpenAI Launches Astra for Law, Enhancing Legal Research Capabilities

OpenAI has introduced Astra for Law, a new AI foundation specifically designed for law firms and legal technology companies, as of September 17, 2026. Astra for Law utilizes a corpus of 230 million URLs, encompassing U.S. case law, statutes, regulations, court rules, and administrative decisions, achieving 99.9% coverage of published U.S. precedential case law through collaboration with the Free Law Project.

In initial testing, Astra for Law demonstrated a 54% correctness check pass rate on legal research questions, significantly outperforming the GPT-6 Astra model, which had a 38.7% pass rate when relying solely on web search. Astra for Law also retrieved 24% more relevant reference cases and 54% more pertinent passages from correct court opinions compared to its predecessor, indicating a 40% relative improvement in performance.

Key legal firms such as Sullivan & Cromwell, Ropes & Gray, and Cooley participated in early testing, with feedback highlighting Astra for Law's strengths in legal research. Niko Grupen, Head of Applied Research at Harvey, noted that the model showed considerable promise across key aspects of legal research. Michael Rubin, Chair of Latham’s AI Strategy Committee, emphasized the importance of deploying AI in environments that require rigorous governance and accountability.

API customers like Harvey and Legora are already building applications on Astra for Law, which could reshape how legal services are delivered. Joel Hron, CTO of Thomson Reuters, expressed optimism about the collaboration with OpenAI, suggesting it could lead to transformative changes in legal workflows. This follows OpenAI's recent initiatives in AI governance and its ongoing commitment to enhancing AI's role in various sectors.

The introduction of Astra for Law marks a significant advancement in legal technology, providing practitioners with a more robust tool for legal research and case analysis.