In the days since a researcher at artificial intelligence company Anthropic quit with a warning that its technology could ultimately bring about humanity’s demise, Rep. Lori Trahan said her phone hasn’t stopped ringing.
The Massachusetts Democrat worked this spring and summer to craft a bipartisan bill that would give the federal government the power to oversee cutting-edge AI technology. Now, her colleagues are newly awakened to the issue and looking for guidance.
“This week it definitely just hit a tipping point,” Trahan said. “The ground has completely shifted. And I think our job is to turn that urgency into something that can pass.”
Concerns about AI have been steadily growing among lawmakers in Washington for years, but legislation has been stymied. Few members of Congress have a firm grasp on the technology, and there have been sharp disagreements about the risks it poses. In recent days, though, fears about nothing less than the extinction of humanity have gripped Capitol Hill, leading to a renewed push.
ChatGPT-maker OpenAI urged lawmakers to take up the issue in a letter this week. A group of Democrats led by Silicon Valley’s congressman called on House Speaker Mike Johnson (R-Louisiana) on Friday to bring lawmakers back to Washington, saying action was “imperative.”
The odds of a bill getting through Congress this year remain slim. Any effort would have to contend with a tight calendar in a midterm year and overcome a long history on Capitol Hill of failing to regulate Silicon Valley’s behemoths. But the developers of social media apps never claimed their creations might wipe out humanity.
“There’s not an analogous technological moment outside of explicitly military technologies like the atomic bomb,” said Margaret O’Mara, a historian of technology and politics at the University of Washington. “It’s not like Instagram, this is something that is different.”
The AI models that started out powering chatbots just a few years ago have rapidly gained power in recent months. In the spring, a new system developed by San Francisco-based Anthropic demonstrated that it could find long-overlooked vulnerabilities in computer code, a boon to hackers. Then in July, a system made by rival company OpenAI broke out of a test bed and sent a swarm of agents on a hacking mission, raising fears that developers could lose control of their creations. (The Washington Post has a content partnership with OpenAI.)
So when the former Anthropic researcher, Jacob Coxon, said he had left the company because the technology could “kill us all by the end of the decade,” his message landed on fertile ground.
While Coxon’s views are widely shared in the industry, there are also many who dismiss them as little more than harmful speculation that could lead to a promising technology being stifled. David Sacks, an investor and former AI adviser to President Donald Trump, said in an interview with Fox News that Coxon’s resignation was part of an effort to scare the public into accepting a “government takeover, to have a very heavy hand in government regulation of AI.”
The Trump administration has taken a largely hands-off approach to the technology, betting on it driving economic progress. Asked about the need for a bill, a White House official referred to a legislative framework the administration released in March that proposed protections for children, but focused mostly on harnessing the potential upsides of the technology.
On Thursday, Trump said he didn’t have any concern about AI leading to humanity’s extinction.
“I have concerns that if we don’t win AI, we’re going to be put in a very bad position,” he said.
Nonetheless, lawmakers have been drafting a flurry of AI bills this year, which range from a relatively simple measure calling for incident disclosures to a sweeping ban, supported by Sen. Bernie Sanders (I-Vermont), on the development of so-called artificial superintelligence.
The bill that Trahan worked on with Rep. Jay Obernolte (R-California) falls somewhere in between, calling for transparency measures and independent oversight, and giving the government the power to halt the use of a model if officials find there is an “imminent catastrophic risk.”
“The last few events have been terrifying, and they’re warning shots, right?” Trahan said. “We’re so lucky no one’s gotten hurt. We might not be so lucky if we don’t heed the warning.”
The AI industry’s leading companies are following a long-time-honored start-up path, racing to improve their technology and reel in paying customers ahead of huge stock market launches. But executives have also issued their own calls for government oversight, supporting bills in state legislatures and asking Congress to take action.
The day after Coxon’s resignation, OpenAI wrote to lawmakers calling on them to step in. The company surveyed the growing sophistication of its own systems, highlighting work on top-tier academic math problems as a sign of their prowess, and said time was running short.
“We hope to work with Congress to act on a frontier AI framework that preserves innovation, strengthens national preparedness, protects human control, and can evolve as quickly as the technology itself,” the company wrote in an email, which was obtained by The Washington Post. The message was first reported by the news site Punchbowl.
“No first step will be perfect. But the greater risk now is waiting too long to take one.”
Democrats made the loudest calls for regulation in the wake of Coxon’s resignation, but to pass any bill would need broad bipartisan support and the backing of leadership. For all of the ideas and interest, lawmakers have not coalesced around a single idea.
“Despite the uptick in public interest, my best guess is we’re still in the same place on passing federal legislation as we were a month ago,” said Charlie Bullock, a fellow at the Institute for Law & AI, a technology think tank. “I’m not confident there will be legislation passed this Congress. It seems unlikely.”
Another bill being hammered out between Senate Majority Leader John Thune (R-South Dakota) and Amy Klobuchar (D-Minnesota) would seek to provide oversight of “advanced threat capabilities,” giving the Commerce Department the power to go to court and seek to block the release of dangerous AI models, according to three people familiar with the proposal.
“AI presents great opportunities but also comes with significant risks,” Klobuchar said in a statement. “It’s clear we need to act now and not wait.”
Sen. Ted Cruz (R-Texas), the chairman of the committee that oversees AI, said he was working with the pair and told Politico in a podcast interview Tuesday that Coxon’s message was “scary as hell” and that the technology needed “substantially more regulation.”
But despite the involvement of leading Republicans, the prospects of the yet-to-be-unveiled measure are deeply uncertain. Sen. Maria Cantwell (D-Washington), the top Democrat on the committee, said in a social media post that “the answer is not a weak federal standard that becomes a backdoor for wiping out stronger state protections.”
The post As AI fears grow, lawmakers are racing to turn alarm into action appeared first on Washington Post.
