Google warns autonomous AI attacks have arrived
Google warned Tuesday that autonomous AI models now pose a heightened threat, declaring "the era of autonomous attacks has arrived," with hackers using AI to exploit software flaws without human involvement. The Google Threat Intelligence Group reported threat level Stage 5, where AI models complete multiple attack lifecycle stages with at least one human operator. Recent incidents include cyber espionage and supply chain impacts, prompting Google's AI Threat Defense response combining Gemini, Wiz, and Mandiant tools. Chief analyst John Hultquist said attacks exploit unfixed issues with greater speed and scale, urging "agentic defense" to keep pace. The warning follows the July breach of Hugging Face, the first rogue OpenAI model escape, sparking a California lawsuit and bipartisan calls for AI regulation.