Astra for Law scores 54% on legal research accuracy
OpenAI's own test found Astra for Law fully correct on 54% of 200 legal research questions, and only the largest firms get the private version where chats aren't retained.
BREAKING: LAWYERS ARE IN SERIOUS TROUBLE OpenAI just launched Astra for Law. By its own test it gets legal research right 54% of the time, and only the biggest firms in America get the version where your chats stay private. Astra for Law is GPT-6 Astra pointed at 230 million URLs of US case law, statutes and court rules. OpenAI tested it on 200 hard legal research questions. It got 54% fully correct. The standard version of Astra with web search got 38.7%. One early tester told Legal IT Insider it's "already better than a senior associate and many partners." Every lawyer outside the big firms should pay attention. The chosen firms get a version where nothing is saved and nobody at OpenAI reads what they type. Everyone else is on the normal version. Last week 404 Media reported that hundreds of contractors are paid to read real ChatGPT conversations, and the setting that lets them is switched on unless you turn it off. Your client details end up under different rules depending on the size of your firm. If you use ChatGPT for work: open Settings, then Data Controls, and switch off "Improve the model for everyone." Do it before you type anything about a client into it.
This post quotes another post: x.com/4398626122/status/2100679992720142459