OpenAI Loads GPT-6 With 230 Million URLs and Isn’t…

OpenAI launched Astra for Law on 17 September, a GPT-6 Astra configuration built around a legal search index spanning more than 230 million URLs of United States case law, statutes, regulations, court rules, and administrative decisions. The release came exactly one week after OpenAI announced ChatGPT for Financial Services with data from LSEG, PitchBook, and Daloopa on 10 September. OpenAI is going profession by profession, building the domain-specific data layer itself, with finance and law arriving one week apart after the company had already applied a similar vertical approach in healthcare.

What Astra for Law Puts on the Table

The product combines GPT-6 Astra with a proprietary legal search index, analytical instructions tuned for legal reasoning, and governance controls designed for confidential client work. The index draws on the Free Law Project's CourtListener database, which covers 99.9% of published United States precedential case law, supplemented by statutes, regulations, and administrative decisions. New sources are added daily, and performance benchmarks tell a clear story. On Vals AI's Legal Research Bench, a 200-question validation set, Astra for Law scored a 54% pass rate compared to 38.7% for GPT-6 Astra with standard web search, at the highest reasoning effort for both systems, representing a 40% relative improvement.The system found 24% more reference cases on case-law questions and retrieved up to 54% more relevant passages from correct opinions. The model appears in ChatGPT as "GPT-6 Astra Law" and will be available on the API as gpt-6-astra-law, though no API pricing or launch date has been disclosed.Access is currently restricted to the Trusted Access programme for eligible law firms, delivered through ChatGPT and Codex. The data policy includes Zero Data Retention (ZDR) on the API and exclusion from human review by default for ChatGPT Enterprise usage, both designed to satisfy the confidentiality requirements that govern attorney-client privilege.

26 Plugins and a Partner List That Includes the Competition

Astra for Law launched with 26 partner-built plugins and nine community plugins offering 47 custom skills. The partner list reads like a directory of the legal technology market: Thomson Reuters, Intapp, iManage, Relativity, Clio, and DeepJudge. Harvey Technologies and Legora are named as API customers building on the model.  Early collaborators include Sullivan and Cromwell, Ropes and Gray, Cooley, Latham and Watkins, and Wachtell Lipton.Joel Hron, Chief Technology Officer at Thomson Reuters, framed the partnership in terms of what raw model access alone cannot deliver: "Legal professionals need more than access to information. They need trusted intelligence, relevant enterprise and matter context, purpose built legal capabilities, and the governance required for high stakes work."Thomson Reuters is integrating HighQ matter context and previewing a CoCounsel Legal connector, positioning its existing workflow tools as the enterprise layer that sits on top of OpenAI's model. The dynamic is worth examining. Harvey, which built its business on proprietary legal AI, is now listed as a customer of the model supplier that is offering a competing product to the same firms Harvey sells to. Sullivan and Cromwell partner John Savva described early internal models as showing "impressive research depth and sensitivity to authority," which suggests the law firms themselves see value in having the model provider compete with their existing vendor stack. ChatGPT for Word also reached general availability on the same day, extending the product into the drafting environment where legal work actually happens.

The Seven-Day Sequence and What It Means for Data Vendors

The FinanceFeeds-relevant story sits in the pattern, not the product. On 10 September, OpenAI launched ChatGPT for Financial Services in partnership with Morgan Stanley and Evercore as design collaborators, integrating LSEG News, PitchBook, Daloopa, Crunchbase, and Quartr as data sources.Users can also connect existing subscriptions to FactSet, S&P Global, Preqin, and Datasite. Exactly one week later, the law received the same treatment with its own domain index and partner ecosystem. The implication for financial data vendors is direct, as OpenAI is not licensing data passively.It is building the workflow layer, the research interface, the cross-source analysis tools, and the client-material generation that LSEG, PitchBook, and Bloomberg sell as their core product. The model supplier is now competing in the application layer where its own data partners operate. The finance launch was powered by GPT-6 Astra with encryption, role-based access controls, and audit-log exports for compliance teams, the same enterprise infrastructure that Astra for Law now uses for attorney-client privilege.The open question is which profession is next. OpenAI has demonstrated it can build a domain-specific data index (230 million URLs for law, multi-source financial data for finance), wrap it in a GPT-6 configuration, ship enterprise governance controls, and recruit the industry's existing vendors as integration partners rather than competitors, all on a timeline measured in days rather than quarters. For data vendors in any regulated profession, the seven-day gap between finance and law is the number that matters. It is the speed at which the model supplier can enter a vertical, and it is faster than most incumbents can respond.