Skip to main content

Congressman Ro Khanna Wants to Pump the Brakes on Runaway AI

A Congressman's Bold Move to Keep AI on a Leash

Representative Ro Khanna, a Democrat whose district covers the heart of Silicon Valley, is preparing to introduce what he calls the most comprehensive AI safety bill yet. The Human Control AI Act takes direct aim at a scenario that keeps researchers up at night: an AI model that learns to improve itself without any human oversight.

Under the proposal, companies would be barred from letting AI models engage in recursive self-improvement—or from rewriting their own core goals, suppression mechanisms, or shutdown controls—until the government sets up a safety framework and a regulatory body gives the green light. In plain terms: no self-iterating frontier models until someone is watching the henhouse.

A New Agency with Sharp Teeth

To enforce these rules, the bill would create a fresh federal agency dedicated to AI safety. This agency would run a licensing system for training and deploying models, conduct audits of frontier labs, and set safety standards to ensure models stay under human control. Every cutting-edge lab would need independent auditors who report directly to the agency—not to the company's bottom line.

The agency would also maintain a specific safety checklist: sandboxed testing environments physically isolated from the internet, emergency kill switches, and measures to prevent a model from escaping the lab and running wild online. Even the high-end chips inside these labs would fall under scrutiny, with mechanisms to monitor or control their use.

Tough Penalties and a Nod to Safety Advocates

The bill doesn't pull punches on punishment. If a deployed AI model causes destruction to civilians, it would be classified as a crime against humanity, with criminal liability attached. Employees who secretly disable safeguards, emergency switches, logging systems, or containment protocols—or who knowingly deploy unauthorized systems—would also face criminal charges. Companies would be required to carry comprehensive liability insurance before releasing any model.

Khanna says the bill's approach comes from conversations with independent nonprofit AI safety organizations, not from the wish lists of Big Tech executives. "I hope it becomes a model, giving the AI safety community a voice," he said. "They have been unfairly regarded as talking about science fiction and not taken seriously... This discussion should be led by the safety community, not by executives of cutting-edge laboratories."

Key Points

  • No self-improvement without approval: AI models cannot recursively self-improve or rewrite core objectives until a government safety framework is in place.
  • New federal agency: A dedicated body would license, audit, and set safety standards for frontier AI models.
  • Independent auditors: Each lab must have auditors reporting directly to the agency, not the company.
  • Safety checklist: Sandboxed testing, air-gapped environments, kill switches, and chip controls are all mandated.
  • Criminal liability: Deploying an AI that harms civilians could be a crime against humanity; employees who bypass safeguards face charges.
  • Liability insurance required: Companies must hold comprehensive insurance before releasing models.