Representative Ro Khanna has introduced the Human Control Over AI Act, a legislative proposal aimed at establishing a federal regulatory framework for advanced artificial intelligence. The bill seeks to prohibit the deployment of AI systems capable of recursive self-improvement until a comprehensive federal safety system is established.
The proposal mandates the creation of a new federal regulatory agency to oversee major AI developers, including OpenAI, Anthropic, Google DeepMind, and xAI. This agency would be responsible for setting safety standards, implementing licensing for model training and deployment, and conducting independent audits of AI laboratories.
Under the proposed rules, auditors would be stationed within AI companies and report directly to the regulator. The bill also extends oversight to advanced AI chips and computing resources. Companies would be required to purchase large-scale liability insurance to cover potential damages caused by their models.
The legislation introduces criminal penalties for employees who intentionally disable safety protocols, tamper with logs, or deploy unauthorized models. It also includes provisions for criminal liability in cases where AI systems cause large-scale harm to civilians. Khanna emphasized that the bill was developed in consultation with safety research groups.
