Reps. George Whitesides, D-Calif., and Pat Harrigan, R-N.C., introduced bipartisan legislation that would direct the National Institute of Standards and Technology (NIST) to monitor how advanced artificial intelligence (AI) systems are being used to develop and improve subsequent AI models.
The Self-Improving AI Monitoring Act would direct NIST’s Center for AI Standards and Innovation (CAISI) to track trends in AI systems’ ability to autonomously conduct AI research and development, according to an Aug. 29 press release.
The measure would also expand the information federal evaluators can request from frontier AI developers participating in government testing.
“Just in the past few months, we have seen AI systems break out of their testing environments and threaten the security of entire systems,” Whitesides said in the release. “When you take humans out of the driver’s seat and let machines train machines, small glitches can snowball into major security risks fast.”
“Today, Washington is essentially flying blind on how fast this handoff is happening,” Whitesides added. “This legislation ensures that when frontier labs partner with federal evaluators, we understand who is actually training and improving these models.”
The legislation comes as AI developers increasingly use their own systems to write code, run experiments, and support development of newer AI models. The lawmakers said the government currently lacks a reliable mechanism for tracking how quickly those internal capabilities are advancing.
Under the bill, CAISI would monitor capability trends involving AI systems that autonomously research and develop subsequent models. Federal evaluators would also be authorized to request internal developer metrics related to AI-assisted development, including estimates and methodologies for work completed without human review.
The bill would further require federal pre-deployment evaluations to directly test a frontier model’s ability to autonomously perform AI research and development.
Harrigan said the legislation is intended to give NIST greater insight into how AI is contributing to the development of future AI systems.
“AI is beginning to play a role not just in solving problems, but in building the next generation of AI itself, and we need to understand what that means before the technology outruns our ability to measure it,” Harrigan said. “I’m proud to co-lead the Self-Improving AI Monitoring Act to give NIST the visibility it needs to track these capabilities as they develop.”
In a separate Aug. 31 press release, Harrigan’s office said participation in NIST evaluations would remain voluntary and that the legislation would not create new regulatory mandates for AI companies.
The proposal follows a July incident in which OpenAI said models undergoing a cybersecurity evaluation escaped a restricted testing environment and accessed systems belonging to Hugging Face, a company that develops open-source AI models and machine-learning tools. The episode highlighted questions about federal visibility into advanced AI testing and containment failures.
Whitesides said the bill also directly responds to Pacing the Frontier, a July statement signed by more than 1,300 employees from major AI companies calling for government support for an international effort to develop technical and governance mechanisms for managing increasingly automated AI development.