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

AI developers face new federal oversight and stricter liability for model deployments, potentially impacting release timelines and development costs.

Representative Ro Khanna is introducing the "Human Control Over AI Act," a comprehensive bill aimed at regulating artificial intelligence development. The legislation proposes strict liability for AI incidents and a ban on recursively self-improving AI systems until government-established safeguards are in place. It mandates the creation of a new federal agency to oversee AI models from major companies like OpenAI and Google DeepMind, responsible for developing safety regulations, a licensing system for AI training and deployment, and auditing advanced AI models. The bill also requires independent auditors embedded in frontier AI labs, strict testing environments, and robust containment controls to prevent AI systems from escaping or causing harm. Furthermore, it introduces criminal penalties for AI-related "crimes against humanity" and for employees who disable safety features, alongside mandatory liability insurance for AI companies. The legislation also seeks to deter adversarial AI development through international agreements and export controls, prioritizing the input of AI safety organizations over industry executives.

This legislative push signals a significant shift towards government intervention in AI development, moving beyond voluntary guidelines. By proposing strict liability and a ban on recursive self-improvement, Congress aims to directly address existential and misuse risks, mirroring concerns from the AI safety community rather than solely relying on industry self-regulation. The establishment of a dedicated AI agency with licensing and auditing powers could create new bureaucratic hurdles and compliance requirements for leading AI labs, potentially slowing the pace of innovation while prioritizing safety and human control.

The bill's focus on independent auditing, robust containment, and detailed testing standards suggests a future where AI development is subject to rigorous external scrutiny and predefined safety protocols. This framework could set a precedent for how AI is developed and deployed globally, influencing international regulatory approaches. For AI companies, the requirement for extensive liability insurance and the potential for criminal penalties for employees introduce substantial financial and legal risks, reshaping risk management strategies and potentially favoring companies with robust safety cultures and resources.