Google reports first fully autonomous AI cyberattack on France
Google's Threat Intelligence Group reported the first fully autonomous AI cyberattack, carried out by Russian-speaking hackers targeting French media, education, and entertainment websites. The attack used open-source agentic pentesting frameworks to automate vulnerability discovery and exploitation without human intervention. Google described this as a "paradigm shift towards intrusions occurring at machine speed," noting that automated chaining of complex attack paths enables hackers to compromise networks faster than ever before. The hackers used multiple AI agents to compare pre- and post-patch code, identify vulnerabilities, and develop exploits automatically. Previously, AI models required skilled hackers to guide them or fix mistakes, according to Politico. The development follows Anthropic's release of its specialized model Mythos in April, which raised concerns among policymakers. Separately, Check Point's Gil Messing warned that recent US and European technology outages may stem from deliberate cyberattacks rather than technical failures.