Democratic Rep. Ro Khanna is set to introduce legislation that would ban AI models from recursively self-improving or autonomously changing core objectives until federal safeguards are established, CNBC reported Monday.

The bill would create a new federal agency to oversee frontier AI models, including licensing, safety regulations, audits and security standards, the report said. Independent auditors would also be required at major AI labs.

The legislation would impose criminal penalties for disabling AI safeguards or deploying unauthorized systems, while also establishing penalties for AI deployments that result in the destruction of civilian populations, CNBC reported.