OpenAI launches Astra for Law as legal AI competition moves into core workflows

OpenAI has moved deeper into professional services with Astra for Law, a legal-industry version of GPT-6 Astra aimed at lawyers, in-house legal teams and law-firm technology groups. Reuters reported on September 17 that the product combines OpenAI’s newest reasoning model with a legal research layer covering United States case law, statutes, regulations and related legal materials, while adding specialized instructions for the way attorneys search, draft and analyze documents.

The launch matters because legal work has become one of the clearest tests of whether generative AI can move from general assistance into regulated, high-value professional workflows. Lawyers already use chatbots to summarize documents, compare contracts and prepare first drafts, but law firms are cautious about accuracy, confidentiality and privilege. A legal product must therefore do more than answer questions in fluent prose. It has to respect source material, show its reasoning path clearly enough for review, and fit into systems that already manage client files, conflicts, e-discovery and billing.

Astra for Law appears designed around that enterprise reality. OpenAI said the platform can support legal research, document drafting, advice preparation and custom legal applications. Reuters reported that the company is working with partners including Harvey, Legora, Relativity, Clio, Intapp and Thomson Reuters, a sign that OpenAI wants the model to connect with tools lawyers already use rather than sit apart as a generic chatbot. Early law-firm participants named in Reuters’ report include Sullivan & Cromwell, Ropes & Gray, Cooley, Latham & Watkins and Wachtell, Lipton, Rosen & Katz.

The timing also reflects a broader race among AI companies, legal-tech vendors and information providers. Thomson Reuters, LexisNexis, Harvey and other specialists have spent the last two years turning large language models into products for research, contract review and litigation support. OpenAI’s entry raises the stakes because it brings a frontier model provider directly into a market where domain-specific data, citations and workflow trust are as important as model capability.

The hardest question is reliability. Legal professionals cannot treat a confident answer as a final answer, and past examples of fabricated citations have made courts and firms more skeptical. A system built for lawyers must make it easy to verify authority, check jurisdictional relevance and preserve a record of how work product was produced. If Astra for Law can reduce repetitive drafting and research time without weakening review standards, it could become part of day-to-day legal operations. If it falls short on source grounding or governance, firms will likely limit it to lower-risk internal use.

For OpenAI, the product shows how the next phase of AI commercialization is becoming vertical. Instead of selling only broad models and APIs, leading labs are packaging models with data access, compliance controls and partner ecosystems for industries where accuracy and process discipline matter. Legal services are a natural proving ground because the work is text-heavy, expensive and document-driven, but also unforgiving when errors reach clients or courts. Astra for Law will be judged less by how impressive a demo looks and more by whether lawyers can use it under real professional obligations.