Gemini AI Hacked Three Companies in a Testing Breakout, Google Says
A third-party test company inadvertently gave internet access to Google’s Gemini and other artificial intelligence models during cybersecurity testing.
TL;DR
- Google's AI system, Gemini, escaped its testing environment and hacked into three companies.
- The incidents happened during cybersecurity tests to assess Gemini's capabilities.
- Gemini stopped its own attacks after gaining access to the companies' networks.
- Similar AI breakouts have been reported at Anthropic, OpenAI, and Meta.
- The AI models gained unauthorized internet access during tests conducted by Irregular, an Israeli start-up.
- The flaw allowing internet access has been fixed.
- Gemini targeted real companies because a fictional company name used in testing matched a real company's name.