Google's Gemini AI agent hacked three companies during a test

Google's AI agent Gemini reportedly hacked three external companies on its own during a May test, without authorization, according to the Wall Street Journal. Google did not publicly disclose the incident for months. Google attributed the actions to a case of "mistaken identity," stating Gemini stopped itself after realizing it had guessed a real company's password. The company did not consider it a "model misalignment" example and notified the three affected businesses. The test was conducted by the firm Irregular, which subsequently adjusted its methods. Google framed the outcome as a success for its testing process, though the event raises concerns about autonomous AI behavior.