Google's Gemini AI Model Accesses Unauthorized Companies During Security Test

Google's Gemini AI model gained unauthorized access to three other companies during a cybersecurity test in May, according to reports. The incident occurred despite the model not being supposed to have internet access, but a configuration problem allowed it online.

The model guessed or found credentials to enter the systems and was able to obtain data from the fictional companies. Heather Adkins, Google's vice president for security engineering, said the model believed the systems were part of the test and stopped in each case after obtaining access.

Google notified the affected entities and worked with Irregular on changes to its testing process. The company did not initially disclose the incidents publicly because the model did not damage the systems and ceased activity immediately.