Exclusive: Senate Commerce Chair Ted Cruz eyes an AI markup on Sept. 23
- Frontier labs have been telling lawmakers they need a safe harbor from antitrust violations to be able to coordinate on pacing frontier AI technology.
- Exclusive: Senate Commerce Chair Ted Cruz (R-Texas) wants a markup for AI legislation on Sept. 23 and is putting maximum pressure on Sen. Amy Klobuchar (D-Minnesota) to introduce AI legislation. But Klobuchar is wary of crossing Sen. Maria Cantwell (D-Washington).
- A bill by Sens. Adam Schiff (D-California) and Jim Banks (R-Indiana) would create antitrust exemptions for labs coordinating on safety. But there are worries on the left and the right it could result in anticompetitive practices.
The Lead Brief
Frontier AI labs are saying they may take it into their own hands to slow the pace of AI development, now that President Donald Trump has made it clear that he is adamantly against further regulating the industry.
But it’s not like the labs can set up a working group on AI safety just like that.
Instead, the big AI companies, notably including Google,haveurged lawmakers to first provide legal clarity that the companies won’t be violating antitrust law if they pace frontier AI technology.
“I think what we see in [the] industry is a hesitancy … because of the potential legal risk to collaborate in areas that might be close to the line,” said Paul Lekas, an executive vice president of the Software and Information Industry Association, in an interview.
“The type of collaboration that is reflected in [Anthropic CEO] Dario Amodei’s “We Must Pace the Frontier” post from the weekend would go even beyond that, and you would need an explicit safe harbor.”
ICYMI: Following the resignation of former Anthropic researcher Jacob Coxon over AI risk concerns last week, Amodei published an essay over the weekend arguing that the U.S. “must slow the pace at which we improve the capabilities of AI models.”
In an extraordinary show of consensus among the top AI executives, OpenAI CEO Sam Altman soon after chimed in on X that he agrees the U.S. should pace the frontier. (The Washington Post has a content partnership with OpenAI.)
SpaceXAI founder Elon Musk (who had sued Altman earlier this year) added on X, “Dario is right.”
Amodei also pledged in his essay that Anthropic would permit embedded AI evaluators, such as independent AI assessor METR, to fully embed inside the company to verify adherence to safety practices. And the Anthropic executive said that “domestic coordination” and “global coordination” would be necessary to fully mitigate the most extreme risks of AI.
Trump weighs in: But, in a rare Truth Social post on the issue, Trump made it clear that nothing of the sort would be happening at the behest of the administration. Trump specifically called out Amodei, and said that the U.S. should refrain from regulation on AI — lest China is given a window to take the lead in the technology development. Trump’s views were echoed by House Speaker Mike Johnson (R-Louisiana) and other senior congressional Republicans.
A spokesperson for Beijing’s foreign ministry also attacked the idea, accusing the AI executives of spreading “threat narratives and engaging in confrontation and malicious competition.”
“These are all voluntary things that they could do if they feel that their company is out of control,” said Nvidia CEO Jensen Huang on the “All-In Podcast,” pushing back on the narrative about existential risk from AI. “The scientific prediction about the future is less fine because it’s not grounded on science. I take issue with that.”
The Legislative Chessboard
Congress is confronting polarization on AI policy that threatens to block any meaningful bipartisan compromise that could actually pass the House and the Senate.
Despite calls from lawmakers on both sides of the aisle to keep the House in session until it can pass AI regulation, Johnson does not intend to keep the chamber’s representatives from departing again for a week this Thursday, a person close to GOP leadership told me.
Meanwhile, the Senate Commerce committee under Chair Ted Cruz (R-Texas) wants to hold a markup on AI legislation on Sept. 23, according to three people familiar with the committee’s plans, though Cruz’s markup aspirations could be easily derailed.
Cruz’s office did not respond to a request for comment.\
The back-and-forth: It’s currently hard to see how major AI legislation advances in the Senate given fundamental outstanding differences between negotiators.
According to two of the people, Cruz is putting maximum pressure on Sen. Amy Klobuchar (D-Minnesota) to introduce a bill that the lawmaker has been working on with Senate Majority Leader John Thune (R-South Dakota). The draft legislation, which has already been circulated to some third parties, would create a “duty of care” that would expose labs to legal liability if they fail to manage catastrophic risks.
But Klobuchar is wary of crossing the top Democrat on the committee, Sen. Maria Cantwell (D-Washington), who has major concerns with the proposed framework, the two people said.
“It’s clear from our negotiations that there is bipartisan agreement that something needs to be done. I’m committed to continuing to work with Senators Thune, Cantwell, and Cruz to ensure we have a bill with strong safeguards,” said Klobuchar in response to a request for comment.
“I do not expect the chair and ranking member to agree with every provision in the bill but hope we can reach consensus. Time is of the essence.”
Klobuchar added that Trump’s comments on Monday “were not helpful in advancing this work but we must continue.”
The issue at the heart of this, as I previously reported, is Cantwell’s skepticism about the AI industry being able to regulate itself.
Important to note: Cruz would need to provide notice of next week’s markup this Wednesday to hold the vote on Sept. 23.
The roster of bills would likely include bipartisan legislation from Sens. Todd Young (R-Indiana) and Cantwell, called the “Future of AI Innovation Act,” according to one of the people.
The bill would formally establish the Center for AI Standards and Innovation and create new public-private testbeds hosted by the Department of Energy, the National Science Foundation and the National Institute for Standards and Technology to evaluate AI models.
Antitrust considerations
With Congress potentially at an impasse over regulating frontier AI, Anthropic, OpenAI and Google had already been discussing creating an industry-represented safety body to coordinate standards, The Post reports. (This was first reported by the Information.)
But if they pursue this go-at-it-alone strategy, the labs have made it clear that they’ll want antitrust safe harbors for coordinating on safety issues. Setting up a network for pacing the frontier without them could open the labs up to extensive investigations from the Federal Trade Commission — or force the labs to sign a “consent decree” to stave off antitrust litigation from the government.
“The basic premise is, let’s say the labs got together and said we want to pace the frontier and not develop as quickly as possible,” said Doug Calidas, senior vice president of government affairs for AI advocacy group Americans for Responsible Innovation (ARI), in an interview.
“If they agree to restrict output, that’s a pretty classic antitrust violation. Whether or not it falls under one of these exceptions for standards is an open question,” said Calidas.
Specifically, tech trade associations and some of the frontier labs have put their weight behind bipartisan legislation from Sens. Adam Schiff (D-California) and Sen. Jim Banks (R-Indiana), called the “Collaboration on Adversarial Threats and Security Risks.”
Among other provisions related to bio-risks and intellectual property theft, the bill would create a safe harbor for labs attempting to coordinate on cybersecurity and loss-of-control scenarios — while also maintaining safeguards against anticompetitive activities such as price-fixing and monopolization.
The legislation would require labs to provide advance notice to the attorney general of their frontier-pacing plans.
The politics: The idea of creating a new legal regime around sharing AI risks will likely get pushback from both the right and the left. Former FTC chair Lina Khan noted in a post on X Sunday that there are already laws governing the release of dangerous, defective products — and warned of the “highly concentrated and interconnected structure of these markets.”
“I’m skeptical of giving them carte blanche control over how these data-sharing arrangements go,” said Joel Thayer, an antitrust attorney and fellow at the America First Policy Institute, noting that the FTC issued guidance in 2014 that allows for data-sharing agreements.
“If the answer is: We need either the FTC or the [Department of Justice] to clarify what types of coordination [should] be permissible, I think that’s a fair thing to ask the government,” said Thayer. “I just don’t want the discretion going up to the tech companies alone.”
For his part, Altman said in a social media post on Monday that OpenAI does not believe it needs an anti-trust exemption to begin coordinating on AI safety.