Mayorkas endorses ‘voluntary’ policies for tackling threats posed by advanced AI models

Press Escape to close the menu.

Concerns are rising that fast-evolving artificial intelligence models, like Anthropic’s Claude Mythos, could be co-opted by hackers to launch cyberattacks at a scale unseen.

Alejandro Mayorkas, the Homeland Security chief during the Biden administration, encouraged the federal government to develop voluntary policies for tech companies to secure rapidly evolving artificial intelligence tools.

“A voluntary paradigm combined with market forces could actually drive progress in safety and security that gets you where you need to go,” Mayorkas said during POLITICO’s Security Summit on Tuesday. “It’s a method of achieving progress, and I think we should hit the pedal on voluntary architecture while this debate is underway.”

He referenced the Biden administration’s work on AI security, which included voluntary standards for deploying frontier models securely, as a blueprint for tackling issues arising from newer models such as Anthropic’s Claude Mythos, which has unprecedented hacking capabilities.