OpenAI Launches Astra for Law With 230 Million URL Legal Index

OpenAI launched Astra for Law on September 17, 2026, a configuration of its GPT-6 Astra model built for legal research, drafting and custom legal applications. The company is not releasing a new base model. Instead, Astra for Law wraps GPT-6 Astra in a dedicated legal search index and a set of instructions tuned for legal analysis and writing. OpenAI is offering the resulting system to law firms and to the software companies that sell tools to them.

The legal search index behind Astra for Law reaches more than 230 million URLs of United States case law, statutes, regulations, court rules and administrative decisions, with sources added daily. The nonprofit Free Law Project's CourtListener database supplies the case law. According to OpenAI, that collection covers more than 99.9% of published US precedential case law. The index is central to the product because it gives the model a curated body of authority rather than the open web.

OpenAI measured the full configuration on 200 US legal research questions drawn from the private validation set of Vals AI's Legal Research Bench. At the highest reasoning effort, Astra for Law passed the benchmark's overall correctness check on 54.0% of questions. GPT-6 Astra using web search alone passed 38.7% of the same questions. The company also said Astra for Law found 24% more reference cases on case-law questions and retrieved up to 54% more relevant passages from correct opinions in a separate audited set.

Access is initially limited. Selected firms reach Astra for Law through a Trusted Access program in ChatGPT and Codex, where it appears in the model picker as GPT-6 Astra Law. An API version named gpt-6-astra-law is planned for later, but OpenAI has attached no launch date or pricing to it. For eligible firms the API comes with zero data retention, and ChatGPT Enterprise usage is excluded from human review by default. Latham & Watkins is working with OpenAI on governance design for information permissions, ethical walls and client instructions. The launch includes 26 partner-built plugins and nine community plugins carrying 47 custom skills.

Key Facts

SiliconANGLE reported on September 17 that OpenAI Group PBC launched Astra for Law as a version of GPT-6 Astra set up for legal research and drafting. The report described the product as a wrapping of the existing model in a legal search index and legal analysis instructions, not as a new foundation model. It also noted that the index spans more than 230 million URLs and that the launch arrived with 26 partner-built plugins connecting ChatGPT to tools such as Relativity, Clio, iManage and DeepJudge. Thomson Reuters is bringing HighQ matter context into ChatGPT and previewing a connector for its CoCounsel Legal product.

MLQ.ai reported on September 18 that OpenAI calls Astra for Law a new foundation while its documentation describes the system as GPT-6 Astra tailored for professional legal work. In ChatGPT the product appears as GPT-6 Astra Law, and the planned API model name is gpt-6-astra-law. The system searches more than 230 million URLs covering US cases, statutes, regulations, court rules and administrative decisions. OpenAI says its work with the Free Law Project brings a CourtListener case-law collection covering more than 99.9% of published US precedential case law into the research experience. The company tested the full configuration on 200 US legal research questions from the private validation set of Vals AI's Legal Research Bench. At the highest reasoning setting, Astra for Law passed the benchmark's overall correctness check on 54% of questions, compared with 38.7% for GPT-6 Astra using web search alone. Those results are company reported and compare Astra for Law with another OpenAI system, rather than with Westlaw, Lexis+ or another commercial legal platform.

Reuters reported on September 18 that OpenAI released a legal industry focused version of GPT-6 Astra designed to help law firms and legal software providers conduct research, draft advice and build custom applications for lawyers. The report placed the launch in a broader race by top AI labs into the legal market. Google last month expanded its Gemini Enterprise offerings for legal professionals, and Anthropic since January has released tools for lawyers using its Claude AI assistant. Reuters also said legal AI vendors including Harvey and Legora will be able to build products on Astra for Law, and that the platform will include integrations with software providers such as Relativity, Clio, Intapp and Thomson Reuters.

Gate News reported on September 17 that the launch came exactly one week after OpenAI announced ChatGPT for Financial Services on September 10. On the 200-question Vals AI Legal Research Bench validation set, Astra for Law scored a 54% pass rate compared with 38.7% for GPT-6 Astra with standard web search, at the highest reasoning effort for both systems, which Gate News described as a 40% relative improvement. The report listed Harvey Technologies and Legora as API customers building on the model and named early collaborators including Sullivan and Cromwell, Ropes and Gray, Cooley, Latham and Watkins, and Wachtell Lipton. Joel Hron, Chief Technology Officer at Thomson Reuters, said legal professionals need trusted intelligence, relevant enterprise and matter context, purpose built legal capabilities, and the governance required for high stakes work. John Savva, a partner at Sullivan and Cromwell, described early internal models as showing impressive research depth and sensitivity to authority. ChatGPT for Word also reached general availability on September 17.

Analysis

What this really means is that OpenAI is competing in legal AI on distribution and workflow integration as much as on raw model capability. Astra for Law is a configuration, not a new base model, and its most important assets are the 230 million URL index and the partner ecosystem around it. By placing a legal model inside ChatGPT and Codex, and by connecting it to tools such as Relativity, Clio, iManage, DeepJudge and Thomson Reuters HighQ, OpenAI is trying to make its model a layer inside existing law firm operations rather than a destination website that lawyers must remember to visit. The 26 partner-built plugins and nine community plugins with 47 custom skills show that OpenAI wants legal software vendors to bring the context and the workflows.

The benchmark design matters. A 54.0% pass rate against 38.7% for web search alone is a 40% relative improvement, but it also means the system missed the overall correctness check on 46% of the 200 questions. The gain from 24% more reference cases and up to 54% more relevant passages suggests that retrieval from a curated legal index is doing real work. Yet the baseline is OpenAI's own web-search configuration, not a commercial legal research platform. That makes the numbers useful for product positioning, but not a decisive proof that Astra for Law beats established legal databases in everyday practice. Law firms will still need to test it against their own matter types, jurisdictions and citation standards.

The bigger picture here is that vertical AI is becoming a land grab across professional services, and governance will decide who wins legal work. OpenAI is limiting the launch to selected firms through Trusted Access, offering zero data retention on the API for eligible firms and excluding ChatGPT Enterprise usage from human review by default. Latham & Watkins is working with OpenAI on information permissions, ethical walls and client instructions. Those controls address the confidentiality rules that govern legal work. If OpenAI can make them reliable, it can become infrastructure for law firms. If it cannot, legal AI vendors such as Harvey and Legora may keep the client relationship even while building on OpenAI models. Google expanded Gemini Enterprise for legal professionals last month, and Anthropic has released tools for lawyers using Claude since January. OpenAI launched ChatGPT for Financial Services on September 10, then followed with Astra for Law on September 17. That sequence shows a company moving industry by industry, using a common model core with specialized indexes, instructions and controls. Incumbents in legal software are mostly integrating rather than resisting: Thomson Reuters, Intapp, iManage, Relativity, Clio and DeepJudge all appear in the partner list. Whether that cooperation lasts depends on how much value OpenAI captures versus how much it leaves to its partners.

Why It Matters

For law firms, Astra for Law changes the baseline for legal research and drafting tools. The CourtListener collection covers more than 99.9% of published US precedential case law, and the full index reaches more than 230 million URLs of case law, statutes, regulations, court rules and administrative decisions. A system that can search that body of authority and retrieve up to 54% more relevant passages on audited case-law sets could reduce time spent finding authority. It could also raise new review duties, because lawyers remain responsible for verifying citations and advice. The product appears in ChatGPT as GPT-6 Astra Law and is planned for the API as gpt-6-astra-law, so the governance terms attached to each surface will matter as much as the retrieval scores.

For the legal technology market, the launch signals that foundation model providers want to own the legal workflow layer. OpenAI launched 26 partner-built plugins and nine community plugins with 47 custom skills. Harvey and Legora are named as API customers building on Astra for Law. That arrangement lets OpenAI gain legal distribution without building every application itself, while partners gain access to GPT-6 Astra and the legal index. The risk for partners is that OpenAI eventually moves closer to the end user. For the wider AI industry, Astra for Law is a test of how vertical configurations are sold and evaluated. The 54.0% versus 38.7% result is company reported and compares two OpenAI systems. The API has no date or pricing, and access is limited to selected firms through Trusted Access. Those constraints mean the market will learn slowly whether the legal index and governance controls translate into trusted daily use. The answer will shape how other regulated industries, from finance to healthcare, adopt similar configurations.

Next Up

The next milestone is the general API release of gpt-6-astra-law. OpenAI has not disclosed a launch date or pricing, so law firms and legal software vendors cannot yet plan broad deployments around it. In the meantime, access remains limited to selected US firms through the Trusted Access program in ChatGPT and Codex. Watch for details on how OpenAI expands eligibility, how it prices legal usage and whether it publishes independent evaluations beyond its own benchmark reporting. Also watch the partner ecosystem. Thomson Reuters is previewing a CoCounsel Legal connector and bringing HighQ matter context into ChatGPT. Harvey Technologies and Legora are building on the model. Early collaborators include Sullivan and Cromwell, Ropes and Gray, Cooley, Latham and Watkins, and Wachtell Lipton. ChatGPT for Word reached general availability on September 17, the same day as Astra for Law. The combination of document-level AI and a legal research index could push OpenAI deeper into the daily work of lawyers, provided the governance controls hold up.