Google thwarts AI-fueled hack attempt

Google announced on Monday it thwarted a cyberattack planned by a criminal group using artificial intelligence to discover and exploit a previously unknown vulnerability in a popular online system administration tool. The company identified the attack as a zero-day exploit, meaning security teams had no prior warning or time to develop a fix. Google traced evidence linking the hackers’ use of an AI large language model—likely not Google’s Gemini or Anthropic’s Claude Mythos—to accelerate the discovery of the flaw, which allowed bypassing two-factor authentication. The incident underscores fears among cybersecurity experts that AI will amplify the speed and scale of cyber threats. John Hultquist, chief analyst at Google’s threat intelligence team, warned that criminal hackers, who operate faster than state-backed groups, now have a significant advantage in identifying and weaponizing vulnerabilities. He described the attack as the beginning of an AI-driven era of exploitation, where hackers can outpace defenses in the race to steal data or deploy ransomware. The Trump administration, which previously rolled back AI regulations introduced by the Biden administration, has since signaled mixed intentions about government oversight. The Commerce Department briefly announced agreements with Google, Microsoft, and Elon Musk’s xAI to evaluate advanced AI models before public release, but the announcement was later removed from its website. Meanwhile, experts like Dean Ball, a senior fellow at the Foundation for American Innovation, acknowledge the need for regulation despite preferring minimal government intervention. Google confirmed it notified the affected company and law enforcement, disrupting the attack before any damage occurred. The company did not disclose the target company’s name or the specific AI model used but ruled out Google’s Gemini or Anthropic’s Claude Mythos. While no direct link to a state-sponsored group was found, Google noted that adversarial actors from China and North Korea have explored similar AI-assisted hacking techniques. The attack follows the release of Anthropic’s Mythos model, which has intensified debates about AI’s potential risks. The incident serves as a stark reminder of the evolving threat landscape, where AI tools could enable cybercriminals to move faster than ever before. As the technology advances, both private companies and governments face pressure to adapt defenses and policies to mitigate emerging risks.