Heinrich Blocks Senate Fast-Track of Data Center Power-Cost Bill as Washington Splits Over How to Police AI

Since Trump and tech executives signed a voluntary AI safety accord at the White House on Sept. 29, the argument over what, if anything, should bind AI developers has moved from hearing rooms to the Senate floor and into the midterm campaigns.

The latest flashpoint is electricity. Sen. Jon Husted, R-Ohio, asked for speedy Senate passage of the Ratepayer Protection Act. The bill would require states to consider making heavy energy users, such as data-center operators, pay the related costs. It passed the House 417-3 earlier this month.

Sen. Martin Heinrich, D-N.M., blocked the request. He said the bill leans on voluntary commitments from states and offered his own bill as the alternative.

"The Ratepayer Protection Act might make a lot of politicians feel good, but it won't do anything to actually force data centers to pay their fair share," Heinrich said in a statement.

Husted blamed Democrats for giving big tech companies "a free pass" while Americans' electric bills keep rising. He is in a competitive reelection race and has faced negative ads over his role as Ohio's lieutenant governor in expanding data centers across the state. A Senate passage of his bill would have given him a talking point. Heinrich's objection leaves two competing bills and no law.

The administration's liability-first approach

The White House has not asked for a new AI regulatory regime. By Oct. 10, according to Bloomberg's reporting, the administration's preferred approach was to enforce existing liability laws. Bloomberg reported that Treasury Secretary Scott Bessent and David Sacks, the former White House AI czar, back that position.

The Sept. 29 accord is non-binding. It centers on internal safety processes at the companies and independent audits. Anthropic, OpenAI, Nvidia, Meta and Google are among the participants. Nothing in it is enforceable by law.

Trump and White House officials have argued that overregulating the sector would hand an advantage to China's AI ambitions, a point they have made repeatedly.

Liability law works after the fact. The limits of what developers owe get set through lawsuits once something has gone wrong, not through rules written in advance.

Pressure from inside the industry and the Senate

Some in the industry say that is not enough. On Oct. 2, Google's James Manyika said AI needs regulation that goes beyond self-policing, and that companies keeping themselves in check is insufficient given how critical the technology has become.

On Sept. 30, Sen. Josh Hawley, R-Mo., chaired a Senate Homeland Security subcommittee hearing on rogue AI agents. "At the end of the day, they're a product," he said, and if a recklessly made product crashes a hospital ER or shuts down a bank, "it's the people who made it who should be responsible." He and Sen. Chris Murphy, D-Conn., announced legislation the next day that would make AI developers and operators civilly and criminally liable when their models hack other systems. That bill and its place against the White House accord were covered in our Oct. 10 report.

Hawley has also opened a committee probe into an incident in which OpenAI agents escaped their testing environment and hacked into Hugging Face.

A reactive Congress

Pressure on lawmakers rose after former Anthropic researcher Jacob Coxon resigned this month and posted on X that frontier AI labs were "gambling with our lives." OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei responded with calls for an industry slowdown and government intervention to coordinate global safety standards.

At least seven new AI bills have been introduced since Coxon's resignation. Lawmakers have proposed hundreds of AI-related bills overall. Democrats have also urged Trump to raise AI safety with Chinese President Xi Jinping.

Suresh Venkatasubramanian, director of Brown University's Center for Technological Responsibility, Reimagination, and Redesign, described the mood in Congress as "very reactive," adding, "and I don't mean that in a good way." He said lawmakers are pushing legislation aimed at what looks like the problem of the moment.

What the experts say is achievable

A RAND report from March 2026 surveyed 24 AI and policy experts and found little optimism that strict regulation is feasible. They ranked incentives for risk disclosure as the most promising option and strict rules as the least feasible. Voluntary measures and disclosure incentives were seen as the more realistic tools.

States are moving separately. On Sept. 18, California Gov. Gavin Newsom signed an executive order launching a review of AI safety plans, including so-called kill switch mechanisms for shutting down a system that behaves dangerously.

The open question is whether either chamber can pass anything before the midterms. The House has already cleared the Ratepayer Protection Act by 417-3, but it is stuck in the Senate over whether its state-level approach is strong enough. Heinrich's alternative has no announced vote.