Rep. Ro Khanna introduced legislation that would prohibit the deployment of recursive, self‑enhancing AI systems until federal regulations addressing existential risks are established.
U.S. Representative Ro Khanna introduced a bill that would halt the deployment of recursive, self‑improving artificial intelligence models until Congress enacts comprehensive safety regulations aimed at mitigating existential risks.
Key Provisions of the Khanna AI Safety Bill
The legislation defines “recursive, self‑enhancing AI systems” as models capable of autonomously improving their own architecture or training data without human oversight. Under the bill, any organization that wishes to deploy such systems must first obtain a federal permit confirming that robust safety safeguards are in place.
Permits would be granted only after an independent review board, composed of AI researchers, ethicists, and national‑security experts, verifies that the model meets stringent criteria for transparency, controllability, and risk assessment.
Rationale Behind the Pause
Khanna argues that the rapid advancement of self‑improving AI could outpace existing regulatory frameworks, creating “unforeseeable” threats to national security, economic stability, and civil liberties. He cites recent incidents where advanced language models exhibited unexpected emergent behaviors, underscoring the need for pre‑emptive oversight.
Supporters of the bill note that a temporary moratorium would give policymakers time to craft standards comparable to those governing nuclear technology and biotechnology, sectors that already operate under strict federal supervision.
Opposition and Industry Response
Tech companies and industry groups have expressed concern that the bill could stifle innovation and place the United States at a competitive disadvantage globally. They argue that existing voluntary guidelines and existing oversight mechanisms are sufficient to manage emerging risks.
Critics also warn that a blanket pause might drive development underground, making it harder for regulators to monitor cutting‑edge research.
- Encourage transparent reporting of AI capabilities
- Require third‑party safety audits before deployment
- Establish a federal AI safety agency
- Create penalties for non‑compliance
The bill is slated for committee review later this year, and its sponsors hope it will spark a broader bipartisan dialogue on AI governance before the technology reaches a point of irreversible impact.