Topic: legal context

GPT-5.2 Tested: 14 Rounds Reveal Serious AI Questions

GPT-5.2 shows only marginal overall improvement, with a surprising regression in coding performance and new, potentially disruptive conversational behaviors like frequently requesting a "go signal" before longer answers. In structured testing, the model excelled in writing, analysis, and creative...