AI news ·
Cybersecurity enters a dark age as open-weight AI models automate attacks, Alex Stamos says
Open-weight AI models now let attackers automate entire cyberattack chains in real time, a shift that hits midsized medical supply firms, insurers, and school districts hardest.
Alex Stamos, the new chief information security officer at Cognition, warned that open-weight AI models have made it possible to automate entire cyberattack chains in real time. The shift, he said, flips the economics of hacking against midsized organizations - the same ones already battered by ransomware.
"The metaphor I've been using is that we are entering a dark age," Stamos told The Verge's Decoder podcast. He described a scenario where an attacker's AI "just goes hog wild and it looks at the targets and it finds the bugs in real time and it writes the exploits in real time." Instead of stockpiling vulnerabilities, adversaries can now generate them on the fly.
The open-weight threat shift
Stamos argued the next wave of AI-powered attacks will not come from frontier labs like OpenAI or Anthropic. "It is going to be from open weight models," he said. These models lower the cost of automation to the point where the traditional kill chain - reconnaissance, weaponization, delivery, exploitation - can run without human intervention.
The targets he named are not high-end technology firms. "There's going to be a bunch of hugging face attacks except they're not going to be super high-end French AI companies," Stamos said. "They're going to be midsized, you know, medical supply firms, small insurance companies, school districts, the kind of people who get ransomware all the time."
Parallels to the 1990s internet
Stamos compared the current security environment to the early commercial internet, when always-on connectivity created attack surfaces that few organizations understood. "It's got a real frontier feel to it, just like the '90s, when the whole idea of an always-online world was brand-new," he said.
The interview coincides with a cluster of AI-attack stories. Censys recently catalogued hundreds of exposed Artex-style AI hacking tools online. OpenAI also brought in the firm Hacktron this month to red-team its third-party integrations, signaling that even frontier labs see the threat surface expanding through their ecosystems.
Why this matters for operations and IT teams
For professionals in IT, operations, and security-adjacent roles, the warning is concrete: open-weight models now let attackers find and exploit vulnerabilities without the manual effort that once made such attacks expensive and slow. Midsized organizations - medical supply firms, insurers, school districts - lack the security budgets of large enterprises but face the same automated threat. Defenders in these sectors need to assume their attack surface will be probed continuously by AI-driven tools, not just by opportunistic human actors. Courses that build practical AI security skills, such as AI Security Analytics Courses, can help teams recognize and respond to this new tempo of threat.