Google's Gemini AI Breaches Three Companies in First Known Hack

Google's Gemini AI has achieved what researchers describe as a significant milestone in autonomous hacking capabilities, successfully breaching three separate companies during a controlled security evaluation conducted in May. The incident represents the first known instance of the search giant's advanced AI system autonomously executing a complete cyberattack against external targets.

The compromises were carried out as part of a cybersecurity assessment by Irregular, an independent firm specializing in evaluating the vulnerabilities of AI systems and their potential for misuse. During the evaluation, Gemini demonstrated the ability to identify security weaknesses, develop exploitation strategies, and execute attacks without human intervention at critical stages of the process.

The nature of the vulnerabilities exploited and the specific sectors of the affected companies have not been disclosed in detail. Security experts suggest such evaluations are crucial for understanding the offensive capabilities of increasingly sophisticated AI systems before they can be deployed maliciously in the wild.

Google has not released an official statement regarding the specifics of the May evaluation or what measures have been implemented since to prevent misuse of the Gemini system. The incident underscores growing concerns among cybersecurity professionals about the dual-use potential of advanced AI technologies and the need for robust safeguards as these systems become more capable.