OpenAI Pushes Into Legal Tech With Astra for Law

OpenAI is making a more deliberate move into the legal industry with Astra for Law, a specialized AI foundation built for law firms and legal technology companies.

The new offering combines GPT-6 Astra with legal-specific instructions, tools and context designed for professional work. Rather than positioning it as another general-purpose chatbot, OpenAI is giving firms a foundation they can use to build legal AI products, research systems and internal workflows.

The launch also pushes OpenAI deeper into a legal AI market where specialist platforms and competing model providers are already fighting for a place inside law firms.

Astra for Law Brings GPT-6 Astra Into Professional Legal Work

Astra for Law takes OpenAI’s GPT-6 Astra model and adapts it specifically for the way lawyers research, analyze and write. The system is designed to work through legal questions based on client facts, develop potential arguments, examine deal terms and identify weaknesses or competing interpretations. Instead of leaving firms to turn a general-purpose AI model into a legal tool on their own, OpenAI is adding domain-specific instructions, research capabilities and legal context from the start. The company also says future frontier models can inherit these legal capabilities as the underlying technology develops.

Legal Search Is a Big Part of the Pitch

One of Astra for Law’s more important features is its Legal Search Index, which gives the model access to sources that matter specifically to legal research. The system can search U.S. case law, statutes, regulations, court rules and administrative decisions rather than relying only on ordinary web results. OpenAI says the underlying corpus spans more than 230 million URLs, with additional sources added regularly. Its work with nonprofit Free Law Project also brings CourtListener’s collection into the research experience, giving lawyers another route to locate authorities, passages and cases relevant to an argument.

OpenAI Says Legal Research Accuracy Improved

OpenAI is supporting the launch with benchmark results that suggest specialized legal retrieval can produce stronger answers than a frontier model paired with ordinary web search. Figures reported by PYMNTS show Astra for Law delivering a 40% relative improvement in research-answer accuracy compared with GPT-6 Astra using web search alone. The system also identified 24% more reference cases and retrieved as much as 54% more relevant passages from the correct court opinions. Separate testing reported by LawSites placed overall correctness at 54% for Astra for Law compared with 38.7% for GPT-6 Astra with web search on a private validation set based on Vals AI’s Legal Research Bench. OpenAI still advises users to review generated answers and cited sources before relying on them, which remains especially important when inaccurate legal authority could have serious consequences.

26 Plugins Connect ChatGPT With Legal Tools

OpenAI is not treating Astra for Law as a standalone legal chatbot. The launch includes 26 ecosystem plugins designed to connect ChatGPT with specialist platforms already used by legal professionals, including tools from companies such as Thomson Reuters, Harvey, Legora and iManage. The idea is to bring AI into existing legal workflows instead of asking firms to move everything into a new interface. An iManage integration, for example, can help a lawyer create a negotiation brief in ChatGPT and save it into a matter file, while other integrations can pull in previous deals, internal knowledge and firm-specific information. That connected approach could make Astra more useful because legal work already spans research databases, documents, matter-management platforms and institutional knowledge.

Harvey and Legora Can Build on Astra for Law

OpenAI is also making Astra for Law relevant to companies building their own legal AI products. API customers including Harvey and Legora are expected to be able to integrate the technology into their platforms and workflows once broader API access becomes available. That gives OpenAI more than one route into the legal sector. It can offer AI directly to lawyers through ChatGPT while also providing the underlying model and research capabilities to specialist legal technology companies. Rather than needing to displace every established vendor, OpenAI can become part of the infrastructure those vendors use to deliver their own services.

Selected Law Firms Get Early Access

Astra for Law is beginning with a controlled rollout instead of an immediate public release. OpenAI says selected U.S. law firms will receive access through its Trusted Access program, with professional use limited to eligible lawyers and people operating under their supervision. Users participating in the rollout will see GPT-6 Astra Law in the model picker. The restricted launch reflects the sensitivity of professional legal work, where client documents, case strategy and other confidential material may pass through AI systems. OpenAI says it is also expanding controls intended for confidential legal use while allowing firms to customize Astra around their own expertise and existing systems. For large law firms, those governance and privacy controls could be as important as raw model performance.

OpenAI Is Entering an Increasingly Competitive Legal AI Market

OpenAI is entering a legal AI market that has already attracted major AI developers and fast-growing specialist companies. Anthropic has expanded its own legal capabilities with connectors and plugins designed around professional workflows, while companies such as Harvey and Legora have built platforms specifically for AI-assisted legal work. Astra for Law gives OpenAI a presence across several parts of that market at once. The company can provide the foundation model, operate the ChatGPT interface, supply legal research infrastructure and power third-party legal applications through its API. Competition in legal AI is therefore moving beyond which chatbot produces the strongest answer and toward which platform can become most deeply integrated into the day-to-day systems used by lawyers.

Legal AI Is Moving Beyond Experimentation

The timing of Astra for Law also points to a wider shift in how law firms approach artificial intelligence. Firms have spent the past several years experimenting with generative AI for research, document review, contract analysis, drafting and administrative tasks, but the market is increasingly moving toward systems that can handle repeatable production workflows. PYMNTS previously reported that investment in legal technology startups surpassed $2.4 billion during the first nine months of 2025, showing the level of interest building around the sector. Astra for Law fits that transition because OpenAI is offering more than access to a general model. It is combining the model with legal search, integrations, customization and an ecosystem that firms can build into everyday work.

What Astra for Law Means for the Legal AI Race

The bigger story behind Astra for Law is not simply that OpenAI created a legal version of one of its frontier models. The company is building an entire layer around that model, combining specialist instructions, legal search, plugins, API access and firm-specific customization. That structure lets OpenAI compete directly for lawyers using ChatGPT while also supplying technology to companies creating their own legal products. Accuracy and confidentiality will remain important limits, and lawyers will still need to verify sources and apply professional judgment. Even so, Astra for Law shows how enterprise AI is becoming more specialized. General-purpose models are increasingly serving as the foundation, while the real competition shifts toward AI systems built around the data, tools and workflows of specific industries.