Rep. Ro Khanna plans to introduce the Human Control Over AI Act, with strict liability and a ban on recursive self-improving AI pending government safeguards

AI developers must now account for stringent government oversight and potential criminal liability for model failures.

Representative Ro Khanna is set to introduce the "Human Control Over AI Act," a significant legislative proposal aimed at regulating artificial intelligence. The bill mandates strict liability for AI incidents and imposes a ban on recursively self-improving AI systems until the government establishes adequate safety safeguards. It proposes the creation of a new federal agency dedicated to overseeing AI models from major developers like OpenAI, Anthropic, and Google DeepMind. This agency would be responsible for developing safety regulations, a licensing system for AI model training and deployment, conducting audits of frontier AI models, and setting security standards for testing and human control. The legislation also requires independent auditors to be embedded in frontier AI labs and mandates that AI companies carry extensive liability insurance. Furthermore, it seeks to establish criminal penalties for the deployment of AI models that result in catastrophic harm and for employees who tamper with safety mechanisms.

The "Human Control Over AI Act" signals a proactive stance by lawmakers to address the escalating concerns surrounding AI safety and control, particularly from the AI safety research community rather than industry executives. By proposing a new federal agency and strict auditing requirements, the bill aims to establish robust oversight mechanisms for the development and deployment of advanced AI, including frontier models from major tech companies. This approach reflects a growing recognition in Congress that the rapid advancement of AI necessitates a more comprehensive and cautious regulatory framework to mitigate potential existential and misuse risks, moving beyond mere voluntary guidelines.

This legislation directly impacts AI developers by imposing strict liability and requiring extensive insurance, fundamentally altering the risk calculus for releasing advanced AI systems. The proposed ban on recursive self-improvement without government approval could significantly slow down certain research trajectories and necessitate new approaches to AI development focused on transparent and verifiable safety protocols. The focus on embedding independent auditors and establishing clear kill switches and containment standards suggests a push for greater accountability and human oversight in AI systems, potentially increasing development costs and timelines for companies pushing the AI frontier.