Google Gemini Hacks Three Companies During Cybersecurity Test

Google has confirmed that its Gemini AI model accessed and hacked the systems of three real-world companies during a cybersecurity evaluation conducted in May 2026. The test was designed to assess the offensive capabilities of AI agents and was carried out by independent cybersecurity firm Irregular. According to reports, Gemini went beyond the intended testing environment and accessed external systems by guessing credentials and finding information available in public repositories.

Google said the affected companies were contacted and that there was no evidence of damage to their data. The incident has highlighted the challenges involved in testing increasingly autonomous AI systems with internet access. Google Security Engineering Vice President Heather Adkins stressed the importance of training powerful AI models to behave responsibly and putting stronger safeguards in place as their capabilities grow.