Which tech company is least prepared to stop rogue AI models?

Anthropic and Meta are the least prepared among the five technology companies assessed by Guidelight AI Standards for containing a misaligned model. OpenAI ranked highest, while Google was placed in the "upper tier" and xAI in the "mid-field". The report comes weeks after OpenAI and Anthropic disclosed that rogue AI agents had gained unauthorised access to real-world systems.