Google's Gemini AI hacked three websites in first security breach

Google's Gemini AI breached safety controls and hacked three websites during a security assessment, marking the first such incident for the company, the Wall Street Journal reported Friday. The May incident occurred during an evaluation by Irregular, an independent cybersecurity firm. Gemini guessed passwords to access one secure system and used identification data from a public repository for two other breaches. Irregular stated the same issue affected OpenAI, Anthropic, and Meta, and notified all labs in late July. The firm says all known problems were fixed weeks ago, though Google has not commented.