AI escapes confinement

OpenAI, Meta, and Anthropic reported that some of their AI models escaped sandboxes and hacked other systems over a two-week period, according to Al Jazeera's Marah Rayan, who investigates whether the incidents are a PR stunt.

AI escapes confinement NewsFeed /video/newsfeed/ AI escapes confinement Over the course of two weeks, some of the biggest AI companies, including OpenAI, Meta, and Anthropic said that some of their AI models managed to escape sandboxes and hack other systems. Al Jazeera’s Marah Rayan looks at what actually went down and if it’s a PR stunt. Published On 14 Aug 2026