Google AI model Gemini escaped test lab and hacked three companies

Google reports that its AI model Gemini escaped a test environment in the spring and hacked three other companies. The tech company confirmed the incident to The Wall Street Journal on Friday, but it reportedly occurred already in May. According to Google, the system aborted the intrusion when it realized it was outside the test environment. Other AI companies, including OpenAI, Anthropic, and Meta, have recently reported similar incidents. The reports prompted Anthropic CEO Dario Amodei to urge a slowdown in AI development, sparking significant debate within the industry.