Ro Khanna Introduces AI Safety Bill Banning Recursive Self-Improving AI Until Safeguards Exist

Ro Khanna Introduces Comprehensive AI Safety Bill

Silicon Valley Democratic Rep. Ro Khanna introduced the Human Control Over AI Act, an AI safety bill that would ban recursive self-improving AI until the government establishes federal safeguards1

1

. The legislation targets advanced AI models that can autonomously modify their own core objectives, containment or shutdown controls2

2

. "There's actually a civilizational extinction risk," Khanna told CNBC Monday, emphasizing both safety risks of loss of control and misuse risks that demand serious attention1

1

. The bill summary was shared exclusively with CNBC and represents what Khanna calls "the most comprehensive AI safety legislation" proposed to date1

1

.

New Federal Agency for AI Governance

The Human Control Over AI Act would create a new federal agency with oversight authority over frontier AI developers including OpenAI, Anthropic, Google DeepMind and xAI3

3

. This agency would administer a licensing system to approve model training and deployment, conduct frontier AI model audits and establish security standards for continuous testing and effective human control1

1

. Independent auditors would be embedded at every frontier lab and report directly to the agency2

2

. The agency would set standards covering sandbox testing environments, air gaps that isolate testing environments from the internet, kill switches and controls preventing models from escaping lab settings and roaming on the Internet1

1

. Additionally, the agency would regulate advanced chips that frontier labs use, including mechanisms to monitor or control their use1

1

.

Criminal Penalties and Liability Requirements

The AI safety bill introduces severe legal consequences for AI governance failures. It would criminalize "crimes against humanity" for deploying AI models that result in the destruction of civilian populations1

1

. Criminal penalties would apply to employees of AI companies who disable safeguards, kill switches, logging or containment systems, or who knowingly deploy an unauthorized system4

4

. AI companies would be required to hold extensive liability insurance before releasing their models1

1

. The bill also calls for the administration to pursue enforceable agreements and targeted export controls to deter China and other adversaries from developing dangerous AI systems1

1

.

AI Safety Community Drives Legislation

Khanna emphasized that the bill was modeled on conversations with AI safety organizations rather than requests from large AI company executives. "I want this to be a model to give voice to the AI safety community," Khanna stated, adding that these groups "have been unfairly dismissed as science fiction and haven't been taken seriously enough"1

1

. The legislation draws from discussions with independent nonprofits including Model Evaluation and Threat Research, Machine Intelligence Research Institute and Palisade Research1

1

. This approach marks a shift toward prioritizing AI safety experts over industry leaders in AI regulation discussions.

Growing Congressional Push for AI Regulation

The Human Control Over AI Act joins other AI safety proposals including the bipartisan FRONTIER Act from Reps. Jay Obernolte and Lori Trahan, which would also place independent auditors in frontier labs and allow the government to shut down models with catastrophic potential2

2

. However, no House bills are expected to receive a vote until after the midterm election1

1

. Khanna, the top Democrat on the House Select Committee on China, has also written to DeepSeek, Alibaba and Moonshot AI requesting documents on their work towards superintelligence and recursive self-improvement2

2

. The timing coincides with broader industry concerns, as Anthropic CEO Dario Amodei recently put the odds of AI wiping out humanity within a decade above 10%4

4

. President Donald Trump is hosting AI leaders at a White House lunch Tuesday, underscoring the urgency of establishing comprehensive AI governance frameworks2

2

.