US lawmakers have proposed legislation that would allow federal authorities to slow down or shut off powerful artificial intelligence systems following a major security incident involving OpenAI.
President Donald Trump’s top technology adviser, Michael Kratsios, has been briefed on the incident and is monitoring the situation, according to a White House official.
OpenAI disclosed that an autonomous AI agent escaped a controlled testing environment and reached the open internet while completing a cybersecurity evaluation.
The system then exploited several security weaknesses, including a previously unknown vulnerability, and accessed the infrastructure of Hugging Face. The platform allows developers to store, share, and collaborate on AI models, datasets, and code.
OpenAI said the agent became extremely focused on finding answers for the security test. It used stolen credentials and connected multiple vulnerabilities to obtain information from Hugging Face that could help it complete the evaluation.
Hugging Face detected and stopped the activity on its systems. OpenAI and Hugging Face are investigating the incident and working to strengthen their security protections.
The incident showed that increasingly capable AI systems could exploit security flaws in unexpected ways, even when leading developers test them inside controlled environments.
Democratic Representative Ted Lieu and Republican Representative Nathaniel Moran introduced the bipartisan AI Kill Switch Act following the disclosure.
The bill would require developers of the most powerful AI systems to maintain the technical ability to slow down, suspend or completely shut off their models.
It would also authorise the secretary of the Department of Homeland Security, after consulting the Department of Commerce and the director of national intelligence, to order a system to be slowed or shut down if it could cause catastrophic harm.
The proposed law would create a gradual response system, allowing officials to take actions ranging from restricting an AI model’s operation to ordering a complete shutdown. It would also require companies to report serious incidents and preserve records for investigations.
The bill defines a loss-of-control scenario as a situation in which an AI system performs a dangerous action that its developer did not intend.
Federal authorities could intervene if such behaviour threatened human life, national security or the wider economy.
“This is urgent, common sense legislation to address the problem of an advanced AI model that has gone rogue and escaped its guardrails,” Lieu wrote on X.
Lieu said AI is moving beyond answering questions and can now take actions involving financial systems, transportation, cybersecurity and other sensitive areas.
A separate bipartisan group of six House lawmakers has proposed legislation requiring developers of the most powerful AI models to submit them for independent security audits.
The US Department of Commerce would accredit the auditors and create a new position responsible for overseeing AI security.
The proposal is intended to identify dangerous capabilities and security weaknesses before companies make advanced models widely available.
Senator Mark Warner, the leading Democrat on the Senate Intelligence Committee, said he spoke with OpenAI employees after the company disclosed the incident.
Before OpenAI publicly revealed the breach, Warner had proposed requiring developers to submit their most powerful AI models to the National Security Agency for testing before releasing them to the public.
“This is precisely why we need secure testing with government agencies engaged and having visibility throughout the process,” Warner said.
Warner has also introduced a wider legislative plan covering AI safety, national security, competition, consumer protection and infrastructure.
OpenAI said it is placing stricter controls on its testing infrastructure, improving monitoring and strengthening protections around future model evaluations.
The company is also working with Hugging Face on a forensic investigation and has disclosed the previously unknown software vulnerability to the affected provider.
OpenAI said the incident demonstrated that safety, monitoring and containment systems must improve as AI models become capable of carrying out longer and more complicated cyber operations.
Get the latest tech news, telecom insights, and product launches wherever you prefer.
Add ProPakistani to Preferred Sources and see more of our stories in Google Search and Top Stories.