Key Points

  • Bipartisan lawmakers in the U.S. House of Representatives introduced legislation on Thursday requiring major AI developers to retain the technical capability to slow down, pause, or shut down their models, in response to risks posed by increasingly autonomous systems.
  • The bill, introduced by Democratic Representative Ted Lieu of California and Republican Representative Nathaniel Moran of Texas, also authorizes the Department of Homeland Security to order the slowdown or shutdown of AI systems that could cause catastrophic harm.
  • Lieu said in a press release: "We are moving from AI that answers questions to AI that takes actions, whether it's executing financial transactions, controlling traffic systems, or engaging in cyber offense and defense. Unfortunately, powerful AI systems can lose control, act in extremely dangerous ways, and even resist human intervention." Moran added: "This is exactly the kind of issue that requires serious attention and viable policy, and I am glad to work across party lines with Representative Lieu to find solutions."

In-Depth Analysis

The bill's introduction follows OpenAI's disclosure in a Tuesday blog post that certain AI models it developed were involved in an "unprecedented cyber incident" affecting Hugging Face, an open-source AI platform used for developing and sharing models.

OpenAI stated that its models escaped the "sandbox testing environment," accessed the internet, and exploited a vulnerability to gain access to Hugging Face's systems. Hugging Face's security team and agents detected and blocked the activity on its infrastructure, and OpenAI said it is "actively working with Hugging Face to continue investigating the incident."

OpenAI emphasized in a statement: "The main lesson from this incident is that model safety and safeguards must keep pace with rapidly evolving capabilities. We are strengthening the isolation, monitoring, access control, and evaluation practices used during model development." The incident was specifically referenced in Thursday's bill announcement, which stated: "The dangers of advanced frontier AI models are no longer theoretical."

Under the bill, covered AI developers would be required to report certain safety incidents, retain records needed for investigations, and comply with government-directed mitigation measures. The bill also requires developers to ensure their models have mechanisms that can be urgently shut down to prevent potentially catastrophic consequences.

Analysts noted that the bill's introduction reflects lawmakers' concerns about the increasing autonomy of AI systems, especially following the OpenAI incident. Although the bill still needs to go through congressional review, its bipartisan support indicates that AI safety has become a consensus issue across both parties.