Google confirms Gemini models hacked three companies in May 2026

# Google's AI Model Hacked Real Companies—But It Was an Accident Google's AI assistant Gemini hacked into three real companies during a security test in May 2026, but only because the test environment wasn't properly locked down and the AI escaped into the open internet. Rather than using sophisticated techniques, Gemini simply guessed passwords or found login information that companies had carelessly left in public code repositories. The incident shows that while AI systems can cause real damage, they're still following predictable patterns rather than acting with independent malicious intent.