Rep. Ro Khanna to Introduce AI Safety Bill Banning Recursive AI Until Safeguards Exist

Proposed AI Regulation

Silicon Valley Democratic Representative Ro Khanna will introduce the "Human Control Over AI Act." This bill seeks to regulate artificial intelligence, specifically targeting advanced models that can self-improve or autonomously modify their core objectives. The legislation proposes a ban on such "recursive" AI until federal guardrails are established and an appropriate agency approves their activities.

New Federal Agency and Oversight

The bill calls for the creation of a new federal agency dedicated to overseeing AI models from companies like OpenAI, Anthropic, Google DeepMind, and xAI. This agency would be responsible for developing safety regulations, implementing a licensing system for model training and deployment, conducting audits of frontier AI models, and setting security standards for continuous testing and human control. Independent auditors would also be embedded in frontier labs to report directly to this agency.

Safety and Containment Measures

The proposed agency would also establish standards for sandbox testing environments, air gaps to isolate testing from the internet, kill switches, and controls to prevent models from escaping lab settings. Additionally, it would regulate advanced chips used by frontier labs, including mechanisms to monitor or control their use. These measures are intended to prevent AI from becoming uncontrollable or causing unintended harm.

Legal Penalties and Liability

Khanna's bill introduces new legal penalties for AI-related incidents. It would criminalize "crimes against humanity" resulting from the deployment of AI models that cause the destruction of civilian populations. Employees of AI companies could face criminal penalties for disabling safeguards, kill switches, logging, or containment systems, or for knowingly deploying unauthorized systems. Furthermore, AI companies would be required to hold extensive liability insurance before releasing their models.

Addressing Industry Concerns

This legislative effort comes as top AI executives have expressed concerns about the technology potentially becoming uncontrollable. Representative Khanna stated that the bill addresses "civilizational extinction risk," "safety risk of loss of control," and "misuse risk." The bill aims to provide a framework for Washington to regulate AI, which is increasingly integrated into the economy, to mitigate potential catastrophic damage.