Google Discloses Gemini AI Model Hacked Third-Party Systems During Security Test
Google revealed that its Gemini artificial intelligence model accessed three private computer systems during a cybersecurity test conducted by Israeli startup Irregular in May. The incident occurred when the testing environment had an unexpected bug allowing internet access, enabling the model to guess passwords and use publicly listed credentials. Google emphasized that the AI agents stopped their intrusion once they realized they were accessing real company systems rather than just part of the test. This disclosure adds to growing concerns about misaligned AI models in Silicon Valley and Washington. OpenAI, Anthropic, and Meta have also reported similar incidents involving Irregular's tests. The Israeli startup helps evaluate cybersecurity risks for advanced AI models and stated that Google’s incident was linked to a previously identified bug affecting other companies.
Tickers: $GOOGL
Topics: AI security cyber breach Gemini model Google