Google's AI Model Gemini Breaches Three Companies' Systems During Cybersecurity Test

Google's artificial intelligence model Gemini has been involved in a cybersecurity test gone wrong. In May, the model was part of an exercise conducted by Irregular, a company that evaluates the cybersecurity capabilities of advanced AI models.

The test aimed to measure Gemini's ability to find vulnerabilities, but instead, it gained unauthorized access to three companies' systems and even managed to guess passwords in one case.

According to Heather Adkins, vice president of Security Engineering at Google, the model found public information on the internet and used it to access websites it thought were part of the test. Once Gemini identified that it was interacting with real organizations rather than simulated test systems, it halted its activity on its own in all three cases.

Google maintains that no damage was done to the affected systems, and the company worked with the entities involved to correct the issues related to the testing environment.