Urgent.News

What's breaking now, across thousands of outlets.

AI

OpenAI launches Astra for Law, combining GPT-6 Astra with a legal search index and instructions for legal analysis and writing, initially for select law firms (OpenAI)

Our most powerful model, configured into a new AI foundation for law. — Explore solutions for law Contact Legal sales

OpenAI has introduced Astra for Law, a specialized version of its GPT-6 Astra model tailored for the legal industry. This model doesn't represent a brand new creation; instead, OpenAI has incorporated a legal search index and instructions for legal analysis to enhance the functionality of GPT-6 Astra. The enhanced model is accessible to law firms and the software companies that cater to this sector.

The index encompasses a vast array of legal resources, including U.S. case law, statutes, regulations, court rules, and administrative decisions, all pulled from over 230 million URLs. The Free Law Project's CourtListener database supplies the case law, while the benchmark used for testing rewards the model's ability to find the correct authority and relevant passages within them.

Astra for Law achieved a 54% accuracy rate on 200 questions from Vals AI's Legal Research Bench, outperforming GPT-6 Astra alone, which could only manage a 38.7% accuracy rate on the same set of questions. On case-law questions, Astra for Law retrieved 24% more reference cases compared to GPT-6 Astra, and up to 54% more of the target passages from correct opinions during an audited set.

Currently, only a limited number of selected firms have access to the model through a Trusted Access program available in ChatGPT and Codex, accessible through the model picker as GPT-6 Astra Law. An API version, gpt-6-astra-law, is expected to follow later without any specified date or pricing. Legal AI companies Harvey Technologies Inc. and Legora AB are among the API customers for this model.

For eligible firms, the API offering ensures zero data retention, and usage of ChatGPT Enterprise is excluded from human review by default. Latham & Watkins is collaborating with OpenAI on governance design for information permissions, ethical walls, and client instructions.

Alongside the model, 26 partner-built plugins have launched, connecting ChatGPT to various legal tools such as Relativity, Clio, iManage, and DeepJudge. Thomson Reuters Corp. is integrating HighQ matter context into ChatGPT and previewing a connector for its CoCounsel Legal product. Another nine plugins were developed by lawyers and legal engineers outside the vendor set, offering 47 custom skills.

ChatGPT for Word was made generally available on the same day as the Astra for Law launch. Joel Hron, OpenAI's chief technology officer, emphasized that while connectivity is important, legal professionals require more than just access to information. They need trustworthy intelligence, relevant enterprise and matter context, purpose-built legal capabilities, and the necessary governance for high-stakes work.

OpenAI has also been developing firm-specific tools on ChatGPT Enterprise, with Sullivan & Cromwell creating an agreement analyzer, Ropes & Gray focusing on deal diligence, and Cooley's tool, GO Public, handling initial public offering preparation.

Written by urgent.news from SiliconANGLE's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

This story

This is one outlet's version. Read the fullest account.

Read the original at openai.com →

More in AI

Anthropic outlines metrics to track AI development at frontier labs: how much AI R&D is done by AI, how well agents are overseen, and how compute is allocated (Anthropic)

AI systems are becoming exponentially more powerful and have begun to automate more of the process of building themselves.

  • Anthropic outlines metrics to assess AI development in frontier labs.
  • Measures include AI involvement in R&D, agent supervision quality, and compute allocation.
  • Independent evaluators will verify safety practices and report incidents.

Jev: The ChatGPT Co-Creator's System One Model Can't Talk

Diogo Almeida’s new model cannot talk. Jev, released in early access on 15 September by TypeSafe AI, returns typed decisions with calibrated confidence: a choice from up to 255 options, a score, or a…

  • Jev is a decision-making model, not a chat model
  • Developed by Diogo Almeida, former OpenAI researcher
  • 193.6 times faster and 444.6 times cheaper than GPT-6 Astra

More from Thursday 17 September →