Without federal AI guardrails in law, bill could chill federal government expert contact with AI companies to address catastrophic risks
WASHINGTON, D.C. – U.S. Senator Maria Cantwell (D-Wash.), Ranking Member of the Senate Committee on Commerce, Science and Transportation, received a commitment from Committee Chair Ted Cruz (R-Texas) to work with her to address concerns that the JAWBONE Act – the goals of which she supports – would undermine the ability of the federal government to keep Americans safe from catastrophic AI-assisted attacks.
“Let me start with what happened with Chairman Carr and ABC and Jimmy Kimmel. Under the current law, Mr. Kimmel had almost no way to vindicate his rights. Chairman Cruz called that dangerous as hell, and he was right. He was right,” Sen. Cantwell said at a Committee markup of several bills. “A speaker should have a real remedy when the government leans on a platform to silence them, and this legislation would provide that remedy.”
But because Congress has not acted on legislation to regulate the AI industry, Sen. Cantwell noted that the bill as written could prevent federal government experts from communicating with AI companies to mitigate serious safety risks.
“There is essentially, right now, no federal law governing AI so under the Chairman’s text, almost any government conversation with an AI company [about] safety could be called coercion,” Sen. Cantwell stated. “For example, if an employee at the Department of Energy asks a company to stop its model from explaining how to enrich uranium, that is content action with no existing federal law behind it. That means that communication could be illegal under this bill, and the employee could be sued.”
Sen. Cantwell offered a substitute amendment to remove AI models. The amendment would also strike the creation of a portal of government communications – a database of government speech; limit liability to decision-making officials in the government; and require a judge to find that any consent judgment is in the public interest.
Earlier this week, Americans for Responsible Innovation also warned Cantwell and Cruz of the unintended consequences of the legislation, writing that: “When a federal official must make a fact-specific determination in the moment, often without a full understanding of the extent of the risk involved and the understanding that she may face civil liability if a court disagrees with her judgment, the human response is to avoid the conversation entirely. If this happens, the American people will be less safe.”
“We know AI can provide a step-by-step instruction on making pathogens like Ebola,” Sen. Cantwell said. “I want scientists at our national labs to work with AI companies to manage these risks. But right now, when we have no federal policy on that, if that scientist or agency said this is dangerous policy, take it down, they could end up being sued under your legislation. So, I ask us to fix this and get it right. I ask us not to have a chilling effect on the communication at an urgent time.”
Although the Cantwell substitute amendment was not agreed to, Chair Cruz agreed to work with Sen. Cantwell to address her concerns.
Video of Sen. Cantwell’s opening remarks is HERE. Video of the JAWBONE debate is HERE. A full transcript is HERE.
###