Welcome to The Hidden Layer. I’m Ian Krietzberg.

It’s been a busy couple days. In my column last week about the intensifying data center debate, one political strategist suggested that Maine Gov. Janet Mills might be worried that her state’s proposed temporary data center ban simply wasn’t good politics. And Mills, who is in a tough Senate primary fight, just proved that prediction right by vetoing the bill on Friday. The Data Center Coalition was thrilled. Also: Yesterday, the Musk v. Altman trial began with jury selection. I’ve got the popcorn ready. This one should be… interesting.

In today’s issue, a close look at Altman’s messaging pivot as the A.I. industry begins to grapple with its self-created image problem. Plus, news and notes on the D.O.J.’s recent regulatory interventions, the White House’s distillation woes, and the latest twists in OpenAI’s open relationship with Microsoft.

Also mentioned in this issue: Dario Amodei, Andy Hall, J.B. Branch, Bernie Sanders, Ron DeSantis, Trump, Michael Kratsios, Chamath Palihapitiya, and more…

Three Things You Should Know…

The Microsoft open relationship saga: On Monday, Microsoft and OpenAI announced another new stage to their partnership that puts a clear upper bound on the hyperscaler’s OpenAI licenses. Under the new terms, OpenAI will continue to pay Microsoft a share of its revenue until 2030—“independent of OpenAI’s technology progress”—and will have the option to grant nonexclusive licenses to its I.P. through 2032. Going forward, Microsoft will remain OpenAI’s “primary” cloud provider, but the trillion-ish-dollar unicorn will be free to work with other clouds as well. Notably, this gets OpenAI out of providing Microsoft with an exclusive I.P. license to its technology, which it had previously agreed to do until it achieved A.G.I., while ostensibly optimizing the company’s $50 billion deal with Amazon.

Amazon C.E.O. Andy Jassy called the announcement “very interesting,” adding, “We’re excited to make OpenAI’s models available directly to customers on Bedrock in the coming weeks. With this, builders will have even more choice to pick the right model for the right job.” Nevertheless, Microsoft remains one of OpenAI’s largest investors; at the time of OpenAI’s conversion into a for-profit enterprise, Microsoft’s stake was valued at roughly $135 billion, or 27 percent of the company.

A MESSAGE FROM OUR SPONSOR

The frontier has moved. Buying tech and AI is easy. Turning it into measurable performance is not. In our work with leading companies, these transformations deliver ~20% EBITDA uplift—but only when they focus on a few domains and change how the work gets done. In Rewired, we show how capabilities turn technology into repeatable, compounding value.

Learn More

The Justice Department v. “woke” A.I.: Earlier this month, xAI sued Colorado in an attempt to kill a 2024 law requiring developers of “high risk” A.I. systems to protect consumers from algorithmic discrimination. xAI’s lawsuit claims that the law would “violate xAI’s constitutional rights and cause irreparable constitutional harm, impose enormous burdens on xAI and the A.I. industry, and substitute Colorado’s political preferences for the national economic and security imperative of American A.I. dominance.”

On Friday, the Department of Justice moved to intervene, arguing that Colorado’s law violated the 14th Amendment. “Laws that require A.I. companies to infect their products with woke D.E.I. ideology are illegal,” Assistant Attorney General Harmeet K. Dhillon, who contested shelter-in-place orders and mask mandates during Covid, said in a statement. “The Justice Department will not stand on the sidelines while states such as Colorado coerce our nation’s technological innovators into producing harmful products that advance a radical, far left worldview at odds with the Constitution.” In fact, algorithmic discrimination is an old, apolitical concept. It’s widely accepted that there are obvious risks to allowing a pattern-matching machine to insinuate itself into human resources departments, say, or the criminal justice system. Nonetheless, the D.O.J., unsurprisingly, wants the law declared invalid.
The U.S. A.I. distillery: Michael Kratsios, director of the White House Office of Science and Technology Policy, warned last week that foreign actors, particularly those based in China, are engaged in a large-scale, “coordinated” campaign to “systematically extract capabilities from American A.I. models.” In reality, model distillation is a pretty standard technique in which a smaller, cheaper model is trained on the outputs of a larger, more advanced model—an approach that extracts greater capabilities across fewer parameters. It’s a method that both OpenAI and Anthropic regularly employ themselves, even though they’ve grown frustrated with their rivals’ use of it during recent months.

OpenAI complained in a February memo to the government that DeepSeek and other Chinese and Russian model developers regularly engage in “adversarial distillation,” an “ongoing effort to free-ride on the capabilities developed by OpenAI and other U.S. frontier labs.” A few weeks later, Anthropic issued a similar report, saying that preventing distillation “attacks” would require a coordinated effort between the industry and the government.

In his warning, Kratsios said that the White House would work more closely with the A.I. industry to help prevent such attacks and explore a “range of measures” to hold responsible the organizations that have been distilling U.S. frontier models. It’s not clear what those measures might look like. Kevin Gosschalk, the C.E.O. of cybersecurity company Arkose Labs, told me that because of model distillation, China is only about six months behind the U.S. “The irony is most of the leading Chinese stuff is all just distillation of the American models,” he said. “The American companies are kind of creating the arms race that’s causing it to get more sophisticated.” A bit of catch-22: As long as U.S. companies keep releasing models, China will keep catching up.

Hallucination of the Week: Gov.AI

Mohammed bin Rashid Al Maktoum, the ruler of Dubai and prime minister of the United Arab Emirates, declared last week that, within two years, “50 percent of [the U.A.E.’s] government sectors, services, and operations will run on Agentic A.I.” How soon until we witness a prompt injection attack against an entire government? As famed software engineer Grady Booch noted, “This will not end well.”

Runner-up: Someone launched a company that uses A.I. to make your A.I.-written emails sound less like A.I. And here I used to think we’d have flying cars by 2025.

And now for the main event…

Love It or Lehane It

With the public continuing to sour on A.I., Sam Altman and his corporate image architect, Chris Lehane, are testing a softer, more human message—less doom and gloom, more uplift and empowerment. Is it too little, too late?

Ian Krietzberg

OpenAI C.E.O. Sam Altman, still shaking off a pair of terrifying attacks on his home in as many weeks, struck a notably softer note during a recent podcast appearance. “There’s a fear with A.I.,” he acknowledged. “Let’s say it makes all this money and does all the work and whatever. Like, what do I do? What’s my kid going to do?” Greg Brockman, the company’s president, jumped in with a similar observation. Most people outside of Silicon Valley don’t understand how they’re going to benefit from A.I., he said. “And I think that’s something that we’re increasingly realizing that we also have to really spell out.”

The industry’s P.R. crisis is, in many ways, a problem of its own creation. For years, Altman and his contemporaries have been warning that the dawn of superintelligence could be an eschatological event with the potential to leave billions jobless—a message that helped A.I. companies raise ungodly sums of capital from investors hoping to squeeze on to the last train leaving the station. Not surprisingly, it’s been less popular with the general public. Poll after poll shows deepening distrust toward the industry, leading to local protests and growing political pushback in Washington.

Now, it appears that OpenAI is trying to sound a little more… human. Earlier this month, the company’s government affairs team rolled out a new policy document, highlighting 20 ideas for “keeping A.I. people-first.” Among the proposals are tax reforms to fund public-safety-net programs, a four-day workweek, and the creation of a public wealth fund to give everyone “a stake in A.I.-driven economic growth.” It’s a markedly different tone than some of OpenAI’s other policy proposals, which focused on playing up America’s technological arms race with China to justify a light-touch regulatory regime.

A MESSAGE FROM OUR SPONSOR

Learn More

Chris Lehane, the ubiquitous political operative and image architect who now heads up global affairs at OpenAI, acknowledged that the company has been working through some of these tensions. “Thus far, the conversation on the social value proposition of A.I. has been overly binary,” he told me. “We don’t think we should just be scaring people,” he added. “We also don’t think we should be telling people that there’s nothing to worry about.”

To that end, this week Altman published a new set of five principles that will guide OpenAI’s work: “democratizing” technological power, “empowering everyone to achieve their goals,” etcetera. Altman painted a picture of a future with “widespread flourishing at a level that is currently difficult to imagine.” None of these new principles will lead OpenAI to stop ramping up its lobbying efforts or staunchly opposing most attempts to regulate the industry. But, Altman wrote, the company is “committed to doing our part to make the future better than the past.” Phew?

Naturally, Lehane dismissed the idea that OpenAI is pivoting its messaging, per se, arguing that it’s simply “the moment” that has changed—that the public is now more receptive to the types of policy proposals that the company is offering. “And no one really has offered them out there yet, and we see a huge opportunity to actually help lead that conversation.”

Public Enemies

I asked Lehane whether he felt any urgency to publicize OpenAI’s new messaging before the politics of A.I. begin to harden—a premise that he politely disputed. “Our urgency is that the nature of the technology itself is going to require transformative policy solutions to optimize for its benefits,” he insisted.

That may be wishful thinking. In 2025, polling from the Pew Research Center found that “near identical” shares of Republicans and Democrats were more concerned than excited about A.I. Only 17 percent of U.S. adults expected A.I. to have a positive impact on the U.S. over the next two decades. Meanwhile, Quinnipiac found that a paltry 5 percent of Americans feel like the people developing A.I. actually represent their interests. Heavy cake.

Meanwhile, public anxiety seems like it’s reached a new and dangerous crescendo, with the attacks on Altman’s house and OpenAI’s headquarters, and an incident in Indianapolis where someone fired 13 shots at the front door of councilman Ron Gibson shortly after the city’s development commission approved a data center project in his district. A note bearing the message “No Data Centers” was left on Gibson’s bullet-riddled front step.

At times, industry luminaries have compounded the problem by publicly chin-stroking about the civilization-destroying possibilities of their technology. In 2014, Elon Musk referred to A.I. as one of humanity’s “biggest existential threat(s).” In 2015, Altman, perhaps partially in jest, said that “A.I. will probably most likely lead to the end of the world, but in the meantime, there’ll be great companies.” Most recently, Dario Amodei, the C.E.O. of Anthropic, put the chances of things going “really, really badly” at 25 percent. All three have suggested that A.I. could devastate the labor market, with Altman regularly calling for a universal basic income, Musk promising that work will soon be “optional,” and Amodei claiming that 50 percent of entry-level white collar jobs will be “disrupted” by 2030.

“I can’t remember another case like this, where the dominant negative narrative on A.I. and the economy is being driven by the C.E.O.s of the companies themselves. I’ve never seen this before,” Andy Hall, a senior fellow at the Hoover Institution, told me a few weeks ago. “I honestly have absolutely no idea what they think they’re accomplishing by going out publicly and constantly talking about how they’re going to put everyone out of work.”

Hall was being somewhat hyperbolic—it’s hardly a total mystery why Altman and Amodei have both pushed this sort of doomer rhetoric. “I get it,” tech investor Chamath Palihapitiya wrote in his 2025 annual letter, “Fundraising requires narrative, and we are building God is a better pitch than we wrote some very clever linear algebra, pirated the internet, and threw a bunch of compute at it.” The problem, Palihapitiya continued, is that A.I. “has to be trusted, because low public trust leads to lower adoption and worse regulations.”

Alas, the tech industry’s track record over the past several decades leaves much to be desired. “People will not trust them until their actions start aligning with what they’re saying,” said J.B. Branch, a policy counsel at Public Citizen. “Even these policy solutions that they’ve come up with—does anyone seriously believe that employers are going to cut the work week? Does anyone seriously believe that we are going to, within the United States, find a way to have universal basic income? These solutions are so devoid of realism that it doesn’t pass a sniff test. It’s just, Let’s say some stuff to appease people, but that’s not going to win anyone over.”

That may be true, but OpenAI, like Anthropic and xAI, does need to do something before public sentiment hardens against the industry. Washington, which is currently an ally, risks turning into an adversary if the most dire predictions about job losses come to pass. At that point, voters won’t be reading Altman’s blog. They’ll be surrounding data centers with pitchforks.

That’s all for today. I’ll see you on Thursday.

Ian

Ace media reporter Dylan Byers brings readers into the C-suite as he chronicles the biggest stories in the industry: the future of cable news in the streaming era, the transformation of legacy publishers, the tech giants remaking the market, and all the egos involved.

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