PlatformsThe briefing
OpenAI says its GPT-6 Astra configuration for legal work pairs an index of more than 230 million US legal sources with partner plugins. Access starts as invitation-only; the API is not out yet.
On September 17, 2026, OpenAI introduced Astra for Law. It pairs its GPT-6 Astra model with a legal search index, plus instructions for legal analysis and writing. OpenAI says the index covers US case law, statutes, regulations, court rules and administrative decisions across more than 230 million URLs, with sources added daily. It says it worked with the Free Law Project, the nonprofit behind CourtListener, whose case-law collection it describes as covering more than 99.9% of published US precedential case law.
Access is limited. OpenAI says Astra for Law starts with selected law firms through Trusted Access in ChatGPT and Codex, and is coming soon to the API. OpenAI gives no price. The index covers US law, and the announcement does not say anything about Europe.
OpenAI also launched 26 plugins from partner companies, such as Relativity, Clio, iManage and Intapp. Thomson Reuters is bringing HighQ matter context in and previewing a CoCounsel Legal connector. There are also 9 community plugins and ChatGPT for Word is now generally available.
For evidence, OpenAI tested 200 US legal research questions from a private validation set of Vals AI's Legal Research Bench. It says Astra for Law passed the correctness check on 54.0% of them, against 38.7% for GPT-6 Astra using web search alone. That is OpenAI's own report of its own test. It does not tell you how the tool will do on your matters. For eligible firms, OpenAI says its API offering includes Zero Data Retention and that ChatGPT Enterprise use is excluded from human review by default.
If you build legal tools, the choice is whether to build on the plugins now or wait for the API. OpenAI says the API is coming soon, so anything you build today inside ChatGPT or Codex may need rework later. Treat the index as a way to find sources, not as the final word. Check every quoted case against a primary source, and record which plugin supplied it. Partner and community plugins are different levels of trust. If you already run legal search, keep your own test set the same and run it through both paths before you change anything.
Eligibility comes first. Trusted Access is for selected firms, so for most firms the product is not available yet, and there is no price to compare with what you pay now. The index is US law only. If you handle client matter data, get the data retention and human review terms in writing instead of relying on a launch post. A lawyer is still responsible for checking the authorities a tool cites.
Ask your OpenAI contact one thing and write down the answer. Is your firm eligible for Trusted Access, on what terms, at what price and with what data retention? While you wait, take ten closed matters and record how long the research took and what it cost. That is the baseline any new tool has to beat.
Source material behind this briefing. Our practical implications are Ashlr’s interpretation.
This briefing was prepared with AI assistance from the linked sources. The analysis may include interpretation; consult the original sources for complete context. Cover artwork is illustrative.
Andy Jassy told analysts that frontier model competition will produce several comparably capable systems. Bedrock already routes to more than 100 models through one API, and most traffic now runs on Trainium.
PlatformsThe briefing
AI-generated illustration
Mistral AI has made the weights of its 1-trillion-parameter multimodal model available under open release, enabling sovereign control and local deployment options.
PlatformsThe briefing
AI-generated illustration
AWS released general availability of training sessions in SageMaker AI serverless model customization. Builders can now write their own post-training loops, including custom loss functions and multi-turn RL, while SageMaker handles GPU allocation, distributed execution, and fault recovery.
