What AI urgency? Industry and leaders split on how to handle advanced tech’s threats

The seeming unanimity among artificial intelligence industry powerhouses about addressing AI’s potential threats to humanity is already splintering.

Four days after OpenAI, Anthropic and Elon Musk locked arms behind a call to slow AI’s advances, a phalanx of industry leaders and lawmakers speaking Wednesday at POLITICO’s Decoded Summit came to no consensus on how to regulate the technology — despite mostly agreeing that it’s a task someone should lead.

Anthropic policy chief Sarah Heck declined to endorse the same proposal for congressionally mandated independent AI monitoring that OpenAI had backed the day before. Progressive Rep. Greg Casar (D-Texas) advocated mandatory limits on the technology and the creation of a federal agency to regulate it, while House Energy and Commerce Chair Brett Guthrie (R-Ky.) said speed is less important than getting the policy right — and the two clashed on whether Americans should care about beating China.

And David Sacks, the former White House AI and crypto czar who wants the U.S. to maintain its leading edge over China above all else, blamed a “media op” for driving panic among elected leaders about AI — a trend he praised President Donald Trump for resisting.

AI developers and policymakers are still largely split on how aggressively to regulate a sector that safety advocates and industry executives have for years warned could pose significant risks to the economy, national security and humanity itself. But now that anxieties over AI have reached a new pitch, leaders in Washington remain paralyzed.

That doesn’t appear to be an issue for Trump, who despite growing pressure on Congress to act has continued to bat away concerns about AI’s theoretical potential to bring about human extinction.

Sacks lavished praise on that approach Wednesday, saying Trump has a unique ability to cut through the panic around AI, which the venture capitalist blamed on a media campaign “to make the American public afraid of AI.”

“The media is pumping out all this fear that makes the public alarmed, and then they poll on that, and then sure enough the polling reflects that sense of alarm,” he said. Sacks added: “So then you get all these politicians saying, ‘Oh, we have to do something, right?’”

Trump “creates the space to then dissect the narrative and find out all the ways in which it is flawed,” Sacks said.

In a series of missives on Truth Social on Monday, Trump repeatedly dismissed growing fears around the technology as a “HOAX,” words that dampened any immediate hope for regulations.

“The only control or ‘guardrails’ that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades,” Trump wrote.

The tech industry recognizes that Americans’ growing opposition toward data centers, anxiety around AI-related economic upheaval and fear of rogue AI agents has shifted the Overton Window of acceptable political dialogue, despite years of calling on Congress for a “light-touch” regulatory approach.

Though the leading AI companies are reupping their public pledges to keep their models safe, the cracks are already beginning to show on the role they want Congress to play.

At the center of the current debate on Capitol Hill is the FRONTIER Act, a bill championed by Reps. Jay Obernolte (R-Calif.) and Lori Trahan (D-Mass.) that would establish Congress’ first federal framework for AI safety but whose future remains murky.

On Tuesday, OpenAI’s top lobbyist Chris Lehane endorsed a key provision of the bill, which would require top AI companies to work with independent safety assessors. But Anthropic’s Heck declined to endorse those same provisions on the Decoded stage Wednesday, even as she insisted the company has “always been for third-party evaluators.”

Casar, the chair of the Congressional Progressive Caucus, is already calling to go much further. Last week, he and Sen. Bernie Sanders (I-Vt.) announced a bill to pause advanced AI development and permanently ban so-called “superintelligence,” in which AI’s abilities outstrip humans’ capacity to counter them.

Hugging Face CEO and co-founder Clem Delangue, whose company found itself the target of an unprecedented cyberattack carried out by rogue AI agents, said during the summit in Washington that tougher liability laws could offer incentives for AI companies to place a priority on cybersecurity for fear of legal peril.

Without it, “you’re going to end up in a world where everyone’s agent is attacking everyone’s with no clear kind of responsibility,” he said.

But he declined to comment on specific legislation and said the current legal framework governing cyber crimes is “working well today,” though it may have to be adapted for an AI-focused future.

Parties remain far apart

Gridlock on Capitol Hill is not for a lack of ideas. But failure to coalesce around legislative solutions to govern AI could turn into a major loss of opportunity for leaders in both parties who have yet to claim the issue as their own. That could be particularly dangerous as the emerging technology becomes a salient issue for voters ahead of the midterms.

Democrats want Congress to act quickly — even if they don’t quite know how.

More than 100 Democrats called on Speaker Mike Johnson to keep the House in session until they pass bipartisan AI legislation. Instead, he cut their time in Washington short on Wednesday, sending lawmakers home one day earlier than expected for a seven-week recess.

Many Republicans, who control gavels in both chambers, haven’t expressed the same urgency to act on AI now.

Guthrie declined Wednesday to commit to timing for a panel vote on a bipartisan proposal to put guardrails on AI — and said it may not even come until next year. Based on current campaign polling, that means it may become an issue for a future Democratic-run House to tackle.

“I’m not saying nothing will happen, but I think a bill of that magnitude needs work,” the Kentucky Republican said.

Casar said at the summit that he met with Anthropic’s Heck on Tuesday morning to urge the company to pause development of its most advanced AI models.

“This isn’t a pause on all AI. It’s a pause on the most dangerous kind of R&D,” he said. But his calls haven’t been echoed by the Democratic caucus.

Legislative details have also tripped up lawmakers.

Representatives from Anthropic, OpenAI, Hugging Face and independent auditing group METR met with Democrats on the House Science, Space and Technology Committee on Tuesday, according to two people granted anonymity to share details of private discussions.

They primarily discussed how to regulate the industry’s most advanced AI systems, but there was no unified consensus on what these guardrails should be, the people said.

In the Senate, negotiations over a bill aimed at addressing the catastrophic risk of AI has hit snags as lawmakers disagree over how or whether the government should vet models before their release.

As fears mount over AI’s advancing capabilities and potential harms, Washington’s partisan decampment has dimmed hopes for legislative action.

But LinkedIn co-founder and Democratic megadonor Reid Hoffman said during the summit that the lack of consensus presents a new opportunity for candidates to tackle the AI issue. Given the public relations crisis the AI industry continues to suffer through, Hoffman sees a narrow lane in which a tech optimist could win back public opinion in 2028.