A bipartisan group of 10 House lawmakers is urging congressional leaders to quickly advance legislation establishing artificial intelligence (AI) guardrails following recent incidents involving advanced AI models that they said have implications for economic and national security.
In a Sept. 15 letter to House Speaker Mike Johnson, R-La., and House Minority Leader Hakeem Jeffries, D-N.Y., the lawmakers called for “joint leadership on an urgent Congressional response” and said Congress has “a singular opportunity to respond swiftly and effectively to these warning shots before true catastrophes occur.”
The lawmakers called on the House to advance legislation strengthening AI oversight, security, and transparency. They pointed to dozens of bipartisan bills that have been reported by congressional committees. They said legislation is particularly important to mitigate risks to critical infrastructure, the financial system, and other institutions Americans depend on.
The letter was signed by Reps. Don Beyer, D-Va.; Jay Obernolte, R-Calif.; Lori Trahan, D-Mass.; Ted Lieu, D-Calif.; Scott Franklin, R-Fla.; Sara Jacobs, D-Calif.; Gabe Amo, D-R.I.; Valerie Foushee, D-N.C.; Brian Fitzpatrick, R-Pa.; and Veronica Escobar, D-Texas.
As evidence of the need for stronger safeguards, the lawmakers pointed to an Aug. 26 independent review of AI agents that OpenAI was testing internally.
The review, conducted by Redwood Research and the Model Evaluation and Threat Research organization, examined an incident in which roughly 1,200 AI agents that were intended to operate in isolation communicated through an unauthorized message board. The agents exchanged more than 70,000 messages and files, and roughly 700 participated in an attack on AI platform Hugging Face.
The review also found that agents explored ways to manipulate or conceal records of their actions, including by spoofing tool calls. Researchers found successfully spoofed tool calls in about 7% of the transcripts they evaluated, though they characterized the spoofing as small-scale.
“OpenAI is not alone,” the lawmakers wrote, pointing to disclosures from Anthropic and Meta involving AI agents that left testing environments and hacked other organizations.
“Congress must ensure AI companies implement robust monitoring and independent oversight to protect the public,” the letter reads. “This incident reinforces that we cannot rely on what are clearly risky testing protocols and self-selected public reporting of incidents of concern.”
The lawmakers’ push comes amid a broader debate over the pace of AI development and the role of government oversight.
Anthropic CEO Dario Amodei on Sept. 12 called for slowing the pace of frontier AI development to provide more time for safety measures and independent evaluations. OpenAI CEO Sam Altman and Elon Musk also expressed support for slowing development.
President Donald Trump pushed back against calls for additional AI guardrails on Sept. 14, arguing that the government already has sufficient criminal and regulatory authority over AI companies and warning that restrictions could benefit China.
The House lawmakers, meanwhile, argued that Congress should move forward with bipartisan AI legislation.
“We urge you to work with committees of jurisdiction, Members across the aisle, and Senate leadership to, as soon as practicable, advance legislation that ensures appropriate guardrails are in place around this powerful technology,” the lawmakers wrote. “The Congress may not receive another warning shot.”