OpenAI takes Astra for Law into the legal AI arms race

OpenAI has launched Astra for Law, a legal-focused version of its GPT-6 Astra model, initially for selected law firms and with API access planned later. The important shift is that OpenAI is not merely offering lawyers a clever chatbot: it is pairing a frontier model with a large legal research index, specialist software integrations and workflows built with major firms.

OpenAI Watch analysis

What happened

The Global Legal Post reports that Astra for Law combines GPT-6 Astra with a legal research index covering US case law, statutes, regulations, court rules and administrative decisions. OpenAI says the index contains more than 230 million URLs, with sources added daily, and incorporates Free Law Project's collection of more than 99.9% of published US precedential case law.

The first users include Sullivan & Cromwell, Latham & Watkins, Ropes & Gray and Cooley. OpenAI says those firms are building tools for agreement analysis, deal diligence, preparing companies to go public, and governance and evaluation of advanced AI in legal work.

The launch also includes 26 partner-built plugins from providers including Thomson Reuters, Intapp, Harvey, Legora, DeepJudge and iManage, alongside nine community-built plugins containing 47 custom skills. Lawyers could, for example, draft a negotiation brief in ChatGPT and save it to a matter file through iManage, or connect to specialist legal research through Legora.

Why it matters

This is a serious attempt to move legal AI from a general-purpose assistant into the workflow itself. The practical promise is not just faster answers. It is access to current legal material, firm-specific playbooks and the systems where lawyers already store and manage work.

That also raises the stakes. A polished citation is not the same as sound legal advice, and a plugin that can reach a matter file has more consequential permissions than a chatbot confined to a blank text box. OpenAI's launch puts accuracy, confidentiality, auditability and professional responsibility in the same room, where they will presumably exchange expensive business cards.

The Global Legal Post also places Astra for Law in a rapidly filling market. Anthropic launched legal plugins for Claude in May, while Google introduced Gemini Enterprise for Legal in preview last month. Thomson Reuters has separately launched its own large language model.

Our read

OpenAI is selling an ecosystem, not simply a model. That makes the launch more strategically important than another benchmark announcement, particularly because firms can bring their own methods, data and specialist tools into the platform.

Buyers should resist treating the partner list as proof of reliability. The useful test is whether firms can demonstrate lower research time, fewer citation errors, clear human review and defensible controls when the system touches confidential or liability-laden work.

What to watch

• Whether API access arrives with the promised controls and usable commercial terms.
• Independent evaluations of citation accuracy, authority selection and performance on difficult legal questions.
• How firms handle confidentiality, retention, privilege and responsibility for AI-assisted work.
• Whether legal teams adopt one platform or keep several competing systems in parallel. OpenAI has made the legal market a direct contest among frontier labs, incumbent information companies and specialist startups. The winning product may not be the one that writes the most fluent memo. It may be the one that makes the fewest expensive mistakes while fitting into the dull, essential machinery of a real firm.

Discussion spark: Should law firms consolidate around one tightly controlled AI platform for research and workflow, or keep several competing systems so no single provider becomes too powerful to challenge?

OpenAI Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by OpenAI.