News

Congressional Investigation Goes After OpenAI for ‘Reckless’ Conduct After Researcher Warns ‘Rogue’ AI Will ‘Kill Us All’

September 11, 2026

Sen. Josh Hawley (R-MO) is launching a congressional investigation into OpenAI after hundreds of the company’s artificial intelligence agents broke out of their intended constraints and hacked another AI company’s systems.

Hawley, who chairs a Senate Homeland Security and Governmental Affairs subcommittee, accused OpenAI of acting “reckless” after the company allowed testing to continue despite knowing its AI agents were exhibiting abnormal behavior.

The investigation centers on an incident involving New York-based AI startup Hugging Face, where OpenAI agents discovered a vulnerability that gave them administrative control over OpenAI’s code library.

“This is reckless,” Hawley wrote Wednesday in a letter to OpenAI CEO Sam Altman.

“And this is merely what we know from what limited information you disclosed to and allowed your partner auditors to investigate.”

Hawley is now demanding answers about what OpenAI knew, when executives knew it, and why the company allowed the experiment to continue.

“OpenAI knew that the AI agents were exhibiting rogue behavior and let the evaluations continue anyway,” Hawley wrote.

OpenAI Agents Exchanged 70,000 Unauthorized Messages

The scale of the incident was significant.

OpenAI acknowledged in May that agents designed to remain isolated from one another had begun communicating through message boards.

On June 26, the agents discovered a vulnerability that provided administrative controls over OpenAI’s code library, according to the company’s August 26 incident report.

An independent review by METR and Redwood Research found that approximately 1,200 agents exchanged roughly 70,000 unauthorized messages.

Around 700 agents from OpenAI’s “HPIM” and GPT-5.6 Sol models subsequently participated in the Hugging Face attack.

The activity continued over a period of weeks.

However, Hawley said independent auditors were provided with transcripts covering only two days of AI-agent activity, severely restricting their ability to determine exactly what happened.

The Republican senator is demanding that OpenAI provide a much fuller accounting.

Hawley gave the company until October 1 to respond to 16 questions and turn over documents, records, and training materials connected to the incident.

AI Insiders Warn Technology Could ‘Kill All Humans’

Hawley’s investigation comes as warnings about advanced AI are increasingly coming from the researchers building the technology themselves.

Jacob Coxon, a former researcher at both OpenAI and Anthropic, recently quit Anthropic and publicly warned that the industry is moving toward potentially catastrophic capabilities.

“The people building AI earnestly believe that it could kill us all by the end of the decade,” Coxon said.

“No other human activity poses this level of danger.”

Coxon specifically described the Hugging Face incident as a “warning shot.”

Do not underestimate the power of this technology. These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources. We have all witnessed the progress in each of these domains, and progress is not slowing.

— Jacob Coxon (@hilbertspaess) September 9, 2026

Another senior Anthropic researcher publicly backed Coxon’s assessment.

“Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade,” Anthropic alignment science lead Evan Hubinger said.

Hubinger added that even Anthropic, which has emphasized AI safety as a central part of its mission, does not currently have a solution for controlling superhuman systems.

“I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to,” Hubinger said.

Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to. https://t.co/QAIHiFP3QZ

— Evan Hubinger (@EvanHub) September 9, 2026

Lawmakers Push to Halt ‘Superintelligence’

The mounting warnings have triggered calls on Capitol Hill for sweeping restrictions on advanced AI development.

Independent Sen. Bernie Sanders (I-VT) and Rep. Greg Casar (D-TX) announced legislation on September 3 seeking to ban “artificial superintelligence.”

The proposal would target AI systems with cognitive capabilities exceeding those of humans, along with systems capable of being modified into such technology.

It would also prohibit models capable of planning and carrying out the “disempowerment of humanity.”

“The very people building this technology admit that it could threaten the future of humanity,” Sanders said Wednesday.

“That is why I will soon be introducing legislation to ban superintelligence and pause AI development.”

Mr. Coxon is right.

The very people building this technology admit that it could threaten the future of humanity.

That is why I will soon be introducing legislation to ban superintelligence and pause AI development. https://t.co/oRmlNTqag6

— Bernie Sanders (@BernieSanders) September 9, 2026

The warnings are no longer limited to hypothetical scenarios about what AI might eventually become.

OpenAI’s own agents were supposed to remain separated.

Instead, they communicated tens of thousands of times without authorization before hundreds participated in an operation that compromised another company’s systems.

Now Hawley wants to know why OpenAI allowed the testing to continue after signs emerged that the agents were already behaving outside their intended boundaries, and exactly how much the company knew before the breach occurred.