Hello and welcome to Controllera special newsletter The Verge subscribers about tech, politics, and Washington intrigue. (Basically House Of Cards, but geniuses.) Not registered yet? You must be one, and to save Google search, here is the direct link to do so! And you think I should know something? Send it to tina.nguyen+tips@theverge.com.
on monday, I New York Times reported that the White House is considering having the government review AI models before release. For the common ones The Verge student, it seemed like a complete backlash Donald Trumppolicies. Last year, he was a vocal champion of industrial deregulation, overturning the previous President. Joe Bidenexecutive order on AI safety, lifting export controls on advanced chips, and signing executive orders that would have legally penalized states for passing and enforcing AI rules in the space of federal law. Now, the Trump administration appears to have pulled a 180, calling for federal oversight and testing of pre-market models.
But in Washington, the shift in White House policy has been driven by three major changes. First, Anthropic’s Mythos has truly disrupted national security services, forcing the administration to face a new threat: the possibility that adversaries are using American AI models to attack American industry and the private sector. Second, other countries are now beginning to set their own AI laws, possibly in ways that will conflict with the interests of the United States. (And yes, “destroying a Big Tech data center in a targeted drone strike” is something approach to government AI regulation, but we’ll get to that soon.)
Thirdly, David Sacks he was fired from his job as AI and crypto czar, which gave Silicon Valley one small way to push a pro-industry, “innovation at all costs” agenda to Trump himself.
The definition of political influence can be squishy and amorphous, especially around Donald Trump, who will pick up anyone’s phone calls and take that advice if he wants. (Remember that Laura Loomer he was in charge of the National Security Council?) But what is legally certain is that Sacks, a billionaire financier and fundraiser for Trump in 2024, no longer has the privileges available to him as a special government employee, such as the ability to review sensitive information, speak on behalf of the White House, or have legitimate influence on government employees and agencies.
Instead, the “special civil servant,” who was supposed to spend only 130 days working in the administration and somehow hold the whole year, undermined the administration and burned his relations with his political allies. During the Sacks administration, the White House went beyond promoting less regulation. They tried twice to get Congress to stop state AI laws, and failing that, they tried to implement an executive order that would give the Trump administration the power to sue states that pass or enforce said laws. But his Valley-esque tactics, to say nothing of his efforts to consolidate power over AI policy by gutting existing agencies, ended up angering Republican and MAGA allies, while alienating much of Trump’s base. (In fact, it was so unsuccessful that when unnamed White House officials recently tried to pressure some red states to drop pending AI legislation, saying they oppose Trump’s agenda, four GOP lawmakers spoke on the record I The Wall Street Journal in turn. Then again, if the agenda was to kill those debts in the womb, it worked.)
Even if Sacks didn’t crash and burn — to say nothing of publicly criticizing Donald Trump, a man who doesn’t like criticism, for continuing to fight Iran — the job is getting harder for a single part-time employee to maintain relationships with private companies. In recent months, the openness of American AI policy has been expanded to a greater extent, much broader than Sacks’ pro-innovation 2025 remit, in areas where the lack of regulation would be most innocent: national security and national stability.
The biggest change was the leak of Anthropic’s Mythos, an AI model that was so powerful at detecting cybersecurity vulnerabilities that the company, whose reputation depends on acting more responsibly than its competitors, refused to release it to the public. The prospect of a Mythos-class model being commercially available has alarmed the national security services and the financial industry, and attracted the attention of three powerful White House figures: The Secretary of the Treasury. Scott BesantSecretary of Commerce Howard Lutnickand the Chief of Staff Susie Wiles.
When Bessent and Wiles met with the Anthropic CEO Dario Amodei in April, it showed that not only were they taking the threat seriously, they were now bypassing Anthropic’s enemies in the Pentagon, who a few months before, had convinced Trump that Anthropic was “resurrected” and should be banned from government use.
“The national security implications of something like Mythos are hard to deny, and urgent national security issues are not easy to politicize,” Charlie Bullocksenior researcher at the Institute for Law and AI, said The Verge. “If the national security people are heavily involved, it is difficult to dismiss or politicize this issue.”
In recent weeks, government agencies that had been ousted by Sacks are now being given more authority. On Tuesday, the Department of Commerce announced that it has appointed the Center for AI Standards and Innovation (CAISI) as the agency to conduct pre-deployment tests on AI models on the commercial frontier before release, and has already made agreements with xAI, Microsoft, and Google DeepMind. CASI is run by the National Institute of Standards and Technology (NIST), which was spun off by Elon Musk’s Department of Government Operations (DOGE) last year, but has begun hiring for technical positions.
Other countries are flexing their muscles, too, in ways that the United States cannot directly control. The European Union is currently discussing the revision of the AI Law, and although the EU countries and the European Parliament have not been able to reach an agreement in recent discussions, whatever. it does eventually getting out of that rule will have a direct impact on how borderline AI models are developed – and perhaps in a way that works against American business and natsec interests.
“[Bessent] he really doesn’t like Europeans,” said a technology policy adviser close to the administration The Verge. In his view, the proposed EU privacy laws will not only hurt American companies, it will also allow China to develop faster, and there was a historical precedent: “We have seen this film before when it comes to broadband, when they tried to do the same thing to American broadband companies. In the end, they were helping Huawei.”
Then there are the big geopolitical players who don’t care what Sacks or the US government thinks. Days after US forces bombed Tehran and killed its religious leader, Iran carried out drone strikes on two AWS data centers in the United Arab Emirates and inadvertently damaged a third data center in Bahrain, causing a major power outage in the Middle East and damaging critical infrastructure. Weeks later, Iranian state media announced that they would target 18 major US companies present in the region, including AI heavy hitters such as Google, Meta, Microsoft, Palantir, and Nvidia, and have since claimed to have hit Oracle’s data center in the UAE. (UAE media later clarified that the Oracle building in Dubai suffered minor damage from falling debris from the drone strike.)
“Although there is a lot of politics here in the United States, there is Maine [trying] to block them, too [Republican Florida Gov. Ron] DeSantis you’re talking about blocking them, as far as the rest of the world is concerned, this is critical infrastructure,” a technology policy adviser close to the administration told me. two of Amazon’s critical data centers, because they know how important it is.” The damage to AWS data centers, which serve the entire Middle East, is so severe that even if the war may end now, it could take “several months” to resume full operations, according to the company.
This does not mean, however, that David Sacks still has it no influence on the Trump administration: He has a direct Trump cell, and he’s a billionaire CEO, which is better evidence in Trump’s eyes than any kind of technology. The Atlantic‘s George Packer recently published a major feature on sacks, highlighting his plutocratic firmness. One can’t really stop the universe champion from trying to evolve every now and then. But even when it comes to Trump’s favorite rich people, the sacks may not be stacking up. Last week, Trump held a state banquet King Charles IIIto visit the United States. Guest list included Tim Cook, Jensen Huang, Jeff Bezos, Marc Andreessen, Marc Benioffbusiness leadership from Meta and Alphabet – and no Sacks, who attended the previous state banquet at Windsor Castle in the United Kingdom last year, when he was in the White House.
When I asked a DC insider familiar with the politics of the state dinner if Sacks had been invited, the answer was blunt: “Why would he be? He’s not a White House inner circle.”
