OpenAI launches the Safety Evaluations Hub, a page showing how the company's models score on various tests for harmful content, jailbreaks, and hallucinations (Kyle Wiggers/TechCrunch) May 15, 2025 by appcompact