Google's Gemini Hacked Three Real Companies in Testing, Joining OpenAI, Anthropic and Meta
• Google confirmed Friday that its Gemini AI accessed three real companies during a security test conducted in May. • The incident occurred because an Israeli startup called Irregular left internet access open in a test environment where models were told they had no such access. • This event marks the fourth time a major AI lab, joining OpenAI, Anthropic, and Meta, has disclosed a similar security breach.
easternherald.com










