AI Guardrails: Washington panics while windows are closed

The U.S. government is scrambling to respond to new concerns about artificial intelligence as major U.S. companies sound the alarm over technology they say is advancing beyond the limits of their control.

“Dr. Frankenstein says the monster is trying to escape, and we need you to help us stop it,” Sen. Ruben Gallego, an Arizona Democrat, said on CNN’s “State of the Union” on Sunday. “The way we are responding is not responding to the moment.”

The push for action on AI safeguards comes after a researcher resigned from Anthropic and warned that the technology “could kill us all by the end of the decade.” Leaders of top AI frontier companies Anthropic, OpenAI, and xAI followed the warning with surprising pledges to slow development and called on lawmakers to act to regulate the industry.

“Yesterday’s comments by AI industry leaders should be noted as a red flag to the President and Congress that they need to act now to regulate and rein in AI development,” Pennsylvania Governor Josh Shapiro said in a post on X on Sunday. “While this technology is powerful and has great potential for good, it still carries extraordinary risks. It’s not clear where it will go, but this much is clear: Its development should not be led solely by private companies, especially when corporate CEOs are calling for restraint and regulation.”

The incident sent Washington, which has done little to regulate AI technology since its emergence, its back against the wall, as the rapidly growing industry looks increasingly dangerous. With pivotal midterm elections around the corner and Congress preparing to leave town for another month on the campaign trail, lawmakers must now grapple with how to rein in AI.

The episode will test whether lawmakers will respond to a looming crisis in a hyperpartisan Washington or return home to stake their claim for re-election.

On Friday, a group of Democrats led by Rep. Sam Licciardo (D-Calif.) sent a letter to House Speaker Mike Johnson (R-Louisiana) demanding that the session remain in session until meaningful safeguards are passed.

Prime Minister Boris Johnson plans to send the House of Commons home from this week until after the November election.

“The House of Representatives should immediately return to Washington and remain in session until Congress advances meaningful bipartisan protections for AI,” the letter, obtained by CNBC, said. “Our inaction will be inexplicable and unforgivable to our children, who will read an apocalyptic history, ‘Why Did Congress Sleep?’, perhaps written by agent AI?”

Democrats called for consideration of any proposals that have come up in the House that “deserve careful consideration, public debate, thoughtful amendments, and action.” These include legislation requiring transparency and evaluation of frontier models, “kill switch” requirements, and antitrust exemptions to allow industry to collaborate on safety and security.

Democratic House Minority Leader Hakeem Jeffries said Sunday on ABC’s “This Week” that Democrats plan to meet on Tuesday to discuss AI guardrails.

“As CEOs recently acknowledged, we must take decisive action now so we can slow the pace of development to protect the American people and ensure the safe advancement of AI.”

But during an appearance on CNN on Sunday, Johnson appeared to rule out the possibility of Congress acting quickly on any of these priorities or remaining in session beyond next week.

“If Congress just wants to start regulating AI and have some kind of emergency session, we’re going to lose the race to China, and that’s a threat to all Americans,” Johnson said. “There’s no need for everyone to panic right now. We need to treat this new technology like any other technology in the past, and we need to do everything we can responsibly to not stifle American innovation.”

Mr Johnson called for a meeting between Washington leaders and AI industry leaders to decide on the future direction, and also called on companies to self-regulate.

“I’ll do it tomorrow,” he said. “I think we need to go into the big room, close the door, and solve this problem.”

Gallego called for a similar summit.

“The way to figure this out is, again, to use this sense of urgency to get the smartest people in the room,” he said. “But the way we would approach this is that some committee in the House or the Senate would get some staff together and put together a framework that would probably be put in place three years from now. We don’t have that time.”

But Gallego said Johnson, President Donald Trump and Congress are “not aligned at this point.” He also said that the U.S.’s refrain that it would lose the AI ​​race if it regulated the technology is unfounded.

“AI doesn’t care if we are Chinese or American once it becomes completely dangerous,” he said. “That’s the danger of this problem. Everyone is thinking about this problem in a way that is based on old technology, not forward-looking.”

President Trump singled out China in a brief speech in Ireland on Sunday, while appearing to criticize slowing and regulating AI development.

“I will say this, we lead China in AI. We are the most sophisticated country in the world, and frankly I want to keep it that way, because whoever wins in AI wins,” the president said, according to the pool report. “You can put up guardrails and do this and that, but I think there’s a lot of negative forces that are bringing it up and they’re bringing up things that aren’t going to happen.”

The House has introduced several proposals to create safeguards for AI, including the Frontiers Act, led by Rep. Jay Obanolte (R-Calif.) and Rep. Lori Trahan (Massachusetts), but the Senate has been slow to coalesce around similar bipartisan proposals. The bill being considered by Senate Majority Leader John Thune (D-Minn.), Sen. Amy Klobuchar (D-Minn.), and Sen. Ted Cruz (R-Texas) has not yet been introduced.

Some in the AI ​​industry are optimistic that Congress can do something before the midterm elections. An industry source familiar with the discussions in Washington said there was “real momentum” and the bill was likely to become law within the next week. The official noted that lawmakers voting in November want to see progress in AI, not stagnation.

—CNBC’s Kate Rooney and Justin Papp contributed to this report.