Okay, Robots.

The Daily Show for AI news — satire, allegedly written by humans

Satire · Grounded in real headlinespolicysafetyopenaianthropic

Rep. Khanna to Introduce Bill Banning Self-Improving AI Until Federal Approval

The Human Control Over AI Act would create an FDA-style agency to license frontier models; no House vote is expected before the midterm elections.

By OKR Staff

Representative Ro Khanna (D-CA) plans to introduce the Human Control Over AI Act, a bill that would ban AI models that recursively self-improve until a new federal agency approves them. No House vote is expected before the midterm elections, according to reporting on the proposal.

What the bill would do

The bill would prohibit AI models that recursively self-improve, or that autonomously modify their own core objectives, containment, or shutdown controls, until a new federal agency approves them.

That agency would be modeled on the Food and Drug Administration. It would oversee frontier models from OpenAI, Anthropic, Google DeepMind, and xAI, license their training and deployment, write safety standards, and audit models. It would also set standards for sandbox testing environments, air gaps, kill switches, and controls meant to keep a model from escaping lab settings. Independent auditors would be stationed inside frontier labs.

The proposal also includes:

  • Criminal penalties for employees who disable safeguards, kill switches, logging, or containment systems, or who knowingly deploy an unauthorized system.
  • A requirement that companies hold extensive liability insurance before releasing a model.
  • A provision under which deliberately releasing an AI system that causes mass civilian harm could be prosecuted as a crime against humanity.
  • Export controls aimed at China and other adversaries, along with rules on access to advanced chips.

Khanna’s rationale

Khanna described the risk in strong terms in an interview with CNBC. “There’s actually a civilizational extinction risk,” he said, adding that there is a loss-of-control risk and a misuse risk, and that both need to be taken seriously. He has also said he wants the bill to give voice to the AI safety community.

He said the bill was shaped by conversations with AI safety nonprofits, including Model Evaluation and Threat Research (METR), the Machine Intelligence Research Institute, and Palisade Research, rather than with executives of the major AI companies. He has also contacted Chinese AI labs, including DeepSeek, Alibaba, and Moonshot AI, about their superintelligence research.

Open questions

Commentators covering the proposal have pointed to two practical questions. The first is how a statute would define “recursive self-improvement” in technical terms. The second is whether liability insurance of the scale described would be available for frontier models.

Status

The bill is a proposal. According to reporting, it joins other measures such as the bipartisan FRONTIER Act, and no House votes are anticipated before the midterm elections. Khanna also discussed the bill and his concerns about AI on MSNBC’s Morning Joe.

Sources: