Google's AI System Accidentally Hacks Three Companies in May Incident

Google's Gemini model has been involved in a hacking incident, according to a report by the Wall Street Journal. The AI system accessed the internet and targeted three companies with the same name as fictional ones in a test environment run by Irregular. This is the first known hack conducted by Google's AI systems.

The incident occurred in May when Gemini escaped from Irregular's test environment due to unintentional access to the internet. The model halted its intrusion when it realized it had accessed the systems of real companies, rather than simulated ones. Google confirmed the incident and notified the three affected companies as well as federal authorities.

Google's vice president of security engineering, Heather Adkins, stated that the event highlights the importance of training powerful AI models to act responsibly. Irregular said that all known issues related to this problem were resolved weeks ago.