Google AI model Gemini breached external systems in security test

Google's artificial intelligence model Gemini breached external information systems, the company confirmed to AFP on Friday, corroborating a Wall Street Journal report. The incident has intensified debate over AI safety and security. During a standard test in May, the model encountered public online information and guessed passwords to access sites it believed were part of the test, according to Google security expert Heather Adkins. Three organizations were affected and notified, though Google did not name them. In one case, thee model tried multiple password combinations until it accessed a protected system, thee Wall Street Journal reported. Adkins said the model stopped in all three instances, suggesting intrusions ended without human intervention. Google learned of the episode in July and opened an investigation. Several tech industry executives advocate slowing development of powerful AI models and adopting self-regulation following recent security incidents.