A MESSAGE FROM OUR SPONSOR

Powering the Energy of Every Day

The electric grid is the backbone of America’s economy. A new Concentric Energy Advisors report highlights how continued investment in reliability and resilience supports homes, schools, offices, and communities across our nation. Watch to learn how investor-owned electric companies are meeting rising demand, maintaining reliability, and keeping customer bills as low as possible as we deliver the energy of every day for 250 million Americans.

After blowing past its full-year A.I. token budget in just four months, Uber has reportedly instituted a monthly token cap of $1,500 per employee, per agentic coding tool. Is the era of out-of-control tokenmaxxing coming to an end? And if so, what does that mean for Anthropic and its 80x surge in revenue? I guess we’ll find out when that S-1 arrives.

And now for the main event…

On Wednesday, OpenAI C.E.O. Sam Altman was back in D.C., taking meetings with White House officials and lawmakers of both parties. As usual, he was there to hype the industry’s safety policies and the rapid progression of frontier models, and to offer his wisdom on various cybersecurity challenges—all with the subtext of persuading Washington not to strangle his company in red tape.

In many ways, Altman was already too late. Just a day earlier, President Trump had signed an executive order requesting that A.I. models be submitted to the government for prerelease vetting—a voluntary submittal, to be sure, but a new imposition nonetheless. Meanwhile, beyond the Beltway, a growing confederation of blue and red states has been loudly advancing regulations of their own.

The latest bureaucratic headache is Illinois’s SB 315, a landmark A.I. safety bill now awaiting the eager signature of Gov. JB Pritzker, the billionaire Democrat and almost certain 2028 presidential candidate. Indeed, while the Illinois bill was partly modeled after similar legislation in California and New York, it goes further than either by requiring every frontier lab with more than $500 million in revenue to submit to annual, third-party audits of their A.I. safety plans. California and New York lawmakers had originally included similar provisions, but faced massive backlash from tech lobbyists. In the end, Govs. Gavin Newsom and Kathy Hochul signed updated versions that stripped out that pesky requirement.

A MESSAGE FROM OUR SPONSOR

Powering the Energy of Every Day

The electric grid is the backbone of America’s economy. A new Concentric Energy Advisors report highlights how continued investment in reliability and resilience supports homes, schools, offices, and communities across our nation. Watch to learn how investor-owned electric companies are meeting rising demand, maintaining reliability, and keeping customer bills as low as possible as we deliver the energy of every day for 250 million Americans.

But the political ground has shifted massively in the six months since those two bills became law. SB 315 had unusually broad, bipartisan support from 93 percent of Illinois’s Democrats, 87 percent of independents, and 88 percent of Republicans, according to the Secure A.I. Project, a nonprofit advocacy group. In the end, both OpenAI and Leading the Future—the pro-industry super PAC that fought the California and New York laws—decided to get on board rather than double down.

The industry has hardly given up on lobbying against regulation. But the strategic retreat did not go unnoticed by insiders tracking the shifting political winds. “I do not think the super PACs are going to get a particularly good return on their investment,” Michael Kleinman, the head of U.S. policy at the Future of Life Institute, told me. “They have had over a year to create a narrative, and they have completely failed,” he said, noting there’s a limit to what money can do when majorities of both parties are skeptical of the industry. “It’s not that they’re powerless, but they are fundamentally fighting a rearguard action.”

For OpenAI, the about-face began in April, when the company threw its public support behind a different Illinois bill, SB 3444, which would have established liability shields for A.I. companies in instances where their tech might cause “critical harms.” It was not exactly a popular position. Less than a month later, after heavy criticism, OpenAI backtracked—testifying that the company did not, in fact, support liability shields and backing SB 315 instead. In its testimony, OpenAI specifically lauded the third-party audits that had once caused so much of its ire, noting that audits can “play an important role in frontier A.I. governance.”

After 315 passed, OpenAI praised the “real bipartisan leadership” behind the bill and its “thoughtful framework for frontier A.I. safety.” The legislation, OpenAI said, demonstrated the “power of reverse federalism—where states lead the way toward establishing a national framework through complementary approaches.” Build American A.I., an arm of Leading the Future, also released a supportive statement, though it noted “potential concerns” about the audit provision. While the group insisted that it believed in a “comprehensive federal framework,” it acknowledged that state legislatures were “helping shape an emerging national model for A.I. governance.”

Industry support for the Illinois bill was not universal. The American Innovators Network, an advocacy group launched less than a year ago to ostensibly represent “little tech” on the political stage, called SB 315 “misguided.” (Curiously, Andreessen Horowitz is funding both Leading the Future and American Innovators Network—an attempt, perhaps, to hedge its bets.) Jeremy Kudon, the executive director of American Innovators Network, told me he’s concerned that too many overlapping, incongruent state laws will create an overly burdensome, inconsistent regulatory “patchwork” that will be impossible for smaller companies to navigate. “What happens if we end up shutting out the next Google, the next YouTube, the next Netflix, because we’re rushing to listen to the two big players?” he said. What happens when a startup crosses the $500 million revenue threshold and suddenly has to “comply with audits that are probably going to be designed for Anthropic and OpenAI”?

Ideally, Kudon argued, “over the next 12 months, you get six to eight more states, maybe 10, that adopt the California SB 53 model.” At that point, he suggested, you essentially end up with a bottom-up, de facto national standard for A.I. Alas, he said, Illinois’s legislators just had to go a little further than California and New York. Now, any company wanting to do business in America’s third-largest metropolitan market will have to comply.

Anyway, the divergence between Leading the Future and the American Innovators Network is just one of the early visible cracks in the battle for A.I. policy influence. On June 1, OpenAI published a blog post in which it sought to distance itself from Leading the Future, which co-founder Greg Brockman and his wife (in addition to Andreessen Horowitz) have funded to the tune of tens of millions of dollars. “We want to be explicit: No outside political group speaks for OpenAI or represents our company’s views,” the post said. Two days later, OpenAI acknowledged that there’s “real momentum right now for A.I. safety policy,” and issued a blueprint for frontier A.I. governance in which it called for transparency, declared it good and necessary for states to function as “laboratories of democracy,” and said that independent, predeployment testing should become a requirement. That’s a pretty dramatic 180 compared to the company’s position even a year ago.

The lesson that Scott Wisor, a policy director at the Secure A.I. Project, has taken from the Illinois saga is simple. “There was a lot of money that came through the Illinois primaries from groups associated with tech companies. All of the legislators who cast a vote in favor of 315 knew that, and they were not scared,” he said. “They looked around, saw the policy, saw the political dynamics, and voted unanimously in favor.” In short, tech money “didn’t change anyone’s mind on how they wanted to cast their vote.” And once tech lobbyists saw which way the wind was blowing, they acted as if they’d been allies all along.

Between Newsom, Hochul, and now Pritzker, the Democratic Party appears to be coalescing around a poll-tested, moderate-ish position on A.I.: implementing guardrails and encouraging innovation. And Republicans, at least so far, aren’t reflexively adopting the opposite position. While there are a few extreme voices on both sides—Sen. Bernie Sanders has called for a total shutdown on data center construction, and Steve Bannon has called A.I. “the most dangerous technology in the history of mankind”—most politicians are somewhere in the middle. And very few are advocating for letting Silicon Valley run wild. “It’s a really untenable position for companies to just say, ‘No rules,’” Wisor said.

Of course, the policy and communications gurus working for these trillion-dollar companies aren’t idiots, either. As OpenAI’s Chris Lehane told me last month, “the moment” has changed—and the hyperscalers need to be active participants in the political conversation. Wisor offered a blunter characterization of this pivot: “Public opinion, concern about your users, concern about your enterprise users and what they’re going to think of you as a company, concern about what your employees are going to think about you as a company—I think it’s all driving the public policy teams at these companies to reevaluate.”

Kleinman, for his part, sees these pivots and policy divergences between industry groups as simply a product of losing. “It’s hard to maintain a coalition when you’re getting beaten back, and so they’re starting to splinter,” he said. “If they were winning, they would all be singing from the same hymn sheet.”

That’s all for today. I’ll see you next week.

Ian

Need help? Review our FAQ page or contact us for assistance. For brand partnerships, email ads@puck.news.

You received this email because you signed up to receive emails from Puck, or as part of your Puck account associated with {{customer.email}}. To stop receiving this newsletter and/or manage all your email preferences, click here.

Puck is published by Heat Media LLC. 107 Greenwich St., New York, NY 10006