At 8AM last Monday, as he prepared for a third marathon day of covering the Senate’s chaotic legislative battle over the Big Beautiful Bill, Steve Bannon’s phone rang. It was Mike Davis, the head of the Article III Project and a lawyer in Donald Trump’s orbit, with an urgent request: he needed to take over the first hour of War Room to raise hell about a ban on states’ AI laws buried in the Big Beautiful Bill. “We have to go in hard on this thing,” he said.

That was a huge ask, Bannon told The Verge. He wasn’t a fan of the AI moratorium, or Big Tech in general, but War Room was built to push its fan base into pressuring Republicans to vote the MAGA way in real time; the Big Beautiful Bill had plenty of things to press them into supporting. And that morning, everyone believed the moratorium issue had been settled: Sen. Marsha Blackburn, a prominent Republican AI hawk who’d initially criticized the ban, had spent the weekend negotiating a compromise with Sen. Ted Cruz that cut the ban to five years and spared certain child safety laws at the state level. If Blackburn was now voting for a moratorium, surely the compromise was acceptable to MAGA populists. But Davis’ plea made him reconsider. “Is it at that stage?” Bannon asked Davis. 

“This compromise is terrible,” Davis responded. “It’s actually worse than the original. We’ve got to kill it.” 

And thus began a 24-hour campaign to reverse what should have been the AI industry’s biggest political win to date. That morning, there seemed to be enough votes in the Republican-held Senate to pass the moratorium, which would have prevented states from writing or enforcing their own laws regulating AI for the next five years, while the federal government figured out a nationwide regulatory framework. (The penalty for breaking the moratorium: states would lose access to a $500 million fund for AI development, which may have been carved out of rural broadband funding.) 

Even if the Democrats were completely unified against it — and considering that it was a Trump-driven bill, they probably would be — the Republicans had the numbers, and could even afford to lose Sens. Josh Hawley and Rand Paul, traditional Big Tech haters who’d voiced their opposition. But by 4AM the next day, after a record-setting 45 rounds of votes and a lobbying meltdown in Washington, virtually the entire GOP had flipped. The bill had passed, but the moratorium had not: 99 out of 100 senators voted for an amendment, sponsored by Blackburn herself, that cut the provision out of the bill. 

According to Republican staffers and conservative tech lobbyists, who were trying desperately to track the votes in real time, this was entirely due to the influence of Bannon and Davis, who spent the entire day battling the moratorium on two fronts. 

In public, they whipped the “War Room posse” into a frenzy, with Bannon, Davis, and other guests railing on-air for hours about the “AI amnesty” that the tech companies were trying to secure for themselves. “3,000 people made 9,000 contacts with their home state senators,” Davis told The Verge afterward. “We lit up their Senate switchboards, all day and all night.” 

Meanwhile, behind the scenes, they were working backchannels — offering support to Blackburn if she chose to back out of the deal, strategizing with staffers and aides, and even going all the way to Trump himself, imploring his team over the phone to hold back and stay silent on this specific issue. 

Getting the AI moratorium killed was victory enough. But getting nearly every Republican senator to cave at 2AM — save for Thom Tillis, who’d just announced his retirement — was something special for the MAGA populists at war against the tech right. 

“I call it the Great Unmasking, because we really saw who are the bitches of Big Tech. And Ted Cruz was the biggest loser in this,” Bannon said. Cruz’s proposal, he claimed, would have forced red states to choose between protecting their citizens from AI, or getting their citizens access to rural internet. “This was absolutely set up to be the cruelest thing you could possibly do. And that’s why he’s nothing but a fucking pimp. And you can quote me.” 

Everyone involved in AI policy, whether they’re lawmakers, interest groups, or industry players, agrees on a few things conceptually. There should be laws regulating AI. There should be laws regulating AI at the federal level. The laws should be thoughtful. The law should not contradict itself. 

But that’s about it; the rest of it is a messy battleground. As with the legislative struggle over digital privacy, there’s a brewing fight over federal preemption — that is, whether federal law overrules and excludes state law on the same matter. Right now, for instance, there’s a piecemeal state-by-state approach to digital privacy, and passage of a federal privacy law is stymied partly by controversy around preemption. Consumer advocacy groups want federal law to be as stringent as what has passed in states like California; otherwise, preemption would roll back protections for Americans. Excluding preemption theoretically means that the federal government provides a floor and states can experiment with increased levels of privacy (think about how the federal minimum wage versus the state minimum wage works). In practice, though, companies that handle data have to comply with an increasingly complex patchwork of privacy regulations. The AI companies are eager not to get into even more of a quagmire, one that is targeted at them specifically.  

The 10-year moratorium was proposed last year by the right-leaning R Street Institute, floated by Cruz during a Senate hearing last month, and added to the House Energy and Commerce Committee’s final draft of the Big Beautiful Bill. 

The justification behind it was this: The best kind of regulation of AI would take place at the federal level and apply broadly across the United States instead of having a piecemeal approach from state to state. Laws take time to write, especially ones on the federal level. During the moratorium, Congress would have time to put together one set of rules, and in the meantime, the AI industry wouldn’t have to tie itself in knots trying to comply with 50 different sets of laws. 

This justification was not well received in many quarters. Piecemeal approaches on all kinds of digital issues (privacy, child safety, and more) have been inconvenient but not existential for industry. But more to the point, 10, even five years is just a very long time to get an extension on Congress’ homework. With the massive impact that AI is already having in all corners of life, the moratorium was a nonstarter for a broad swath of Americans. It was no surprise that there were objections from consumer protection groups and state legislators already trying to write their own laws in a regulatory vacuum. But the moratorium also happened to draw heavy Republican opposition, a phenomenon rarely seen these days in a party loyal to Trump: in the run-up to the voting period, 37 state attorneys general and 17 governors sent letters to Senate Republican leaders, urging them to get rid of the moratorium and protect states’ rights.

The MAGA mediasphere did not like it, either, and they’d glommed onto the issue in early June when Rep. Marjorie Taylor Greene (R-GA) admitted something embarrassing: she hadn’t known the moratorium was in the 940-page bill. “I am adamantly OPPOSED to this and it is a violation of state rights and I would have voted NO if I had known this was in there,” she posted on X on June 3rd, a full two weeks after the tax bill passed the House, arguing that it was too dangerous to “tie states’ hands” for the next 10 years. “When the OBBB comes back to the House for approval after Senate changes,” she threatened, “I will not vote for it with this in it.” Considering the GOP holds the House by a historically slim 8-vote margin, losing Greene would throw the OBBB’s passage in jeopardy.

However, it wasn’t as simple as deleting the offending clauses. The moratorium’s proponents had pulled some procedural shenanigans by locking it into the House’s version of the Big Beautiful Bill, and to remove it, the Senate needed to pass an amendment that explicitly cut out the language. But they only had until July 4th to pass the bill, a deadline imposed by the White House, and there were already too many amendments piling up in the upcoming “vote-a-rama” — a Senate procedure that’s also a unique form of psychological torture. 

Senators are allowed to propose an unlimited number of amendments to any budget-related bill that’s made it to the floor — either to make a political point, or occasionally, to pass an actual piece of legislation — and force their colleagues to consider, debate, and vote on every single one, even if it takes endless days to do so. Ideally, it would crush their opponents’ will to live, or at least get them to change their votes. (The sleep deprivation torture had already started: the Democrats had used a procedure that required the Senate clerks to read the entire 940-page bill before they could start voting on amendments. It took 16 hours.) 

Blackburn had initially been the Republicans’ most outspoken critic of the moratorium, but once she made her deal with Cruz, a ban — reduced down to five years, instead of 10 — was almost certain, even though Democrats like Sen. Maria Cantwell (D-WA) were now submitting amendments to remove it. As long as Republicans held the party line, the moratorium was in. Sure, Greene was a potential spoiler in the House, but Trump had just forced Tillis into retirement after the North Carolina Republican criticized the OBBB’s Medicare hikes. “I don’t want to suggest that that’s going to happen to Greene,” Adam Thierer, a fellow at the conservative R Street Institute and one of the original proponents of the moratorium, told The Verge on Monday afternoon, “but I think there’s a lot of Republicans that live in fear of being primaried based upon opposition to certain Trump priorities.”

Bannon and Davis weren’t having it, however. The moment that Davis went off air on Monday morning, Bannon ended up on a call with Blackburn. “I said, ‘Listen, our audience loves you and we have your back,’” said Bannon. “‘We will make as many phone calls, and send as many text messages [as you need]. You do what you have to do, but don’t think you have to compromise.’” (Blackburn’s office did not initially respond to a request for comment. After publication, Audrey Cook, a spokesperson for Sen. Blackburn, disputed the characterization that Bannon had persuaded Blackburn to back out of the compromise with Sen. Cruz.)

By 2PM, the Republican-connected tech lobbyists had started to fret. “Sounds like Sen. Blackburn will be offering an amendment to strip,” one lobbyist texted The Verge. “Thought we were good on the amendment, but who knows.” 

Thierer, too, had no idea what was happening in the Senate, which was wild because the R Street Institute had virtually written the blueprint for the moratorium in 2024. “At this point, every couple of hours, the situation on the ground appears to change.
And it’s like you have to date-stamp your thoughts on a millisecond-by-millisecond basis because the roller coaster continues,” he said. “I’ve almost given up trying to write anything that’s going to be fresh longer than 24 hours on this.”

Around 6PM, Bannon made a stunning announcement on War Room: Blackburn had decided to withdraw her support from the moratorium and would vote against it. For a while, it wasn’t clear whether Bannon was correct: a Republican aide told reporters that Blackburn was still in, and Blackburn hadn’t made a statement on it yet. But Bannon had been on the phone with her right before he dropped that bombshell. “I said it because she told me she was a no,” he told The Verge.

It took two hours for Blackburn to officially confirm she was backing out of the compromise, sending out a statement at 8PM saying that the Cruz provision was “not acceptable” and “could allow Big Tech to continue to exploit kids, creators, and conservatives.” Then, at 9PM, she made it clear that she really opposed the moratorium: she officially filed her own amendment that would strip the language from the bill entirely, with Cantwell as a cosponsor. 

With Blackburn, Hawley, and Paul joining the Democrats, the Senate was unofficially deadlocked at 50-50. When Sen. Susan Collins (R-ME), one of the more moderate members of the GOP caucus, shockingly signed onto the amendment, the moratorium was probably dead, though Cruz was still upbeat. “The night is young,” he told a Punchbowl reporter, as the vote-a-rama period hit 10PM with no end in sight. 

Davis, in the meantime, had been making calls to the White House. Trump had reportedly given his blessing to the Cruz-Blackburn five-year compromise over the weekend. The pro-moratorium interest groups were hoping that he’d publicly back the bill, which would have very likely whipped MAGA loyalists in line. But Davis was a formidable foe — he was close enough to Trump himself to be able to text the administration. 

“I told the key people in the Trump administration not to support this,” Davis told The Verge. “These AI oligarchs hate us and now they want to steal every copyright in the world, harm kids for profit, and cancel conservatives and others with whom they disagree. Why the hell do we want to give them 10 years of AI amnesty?”

The rumor had been that Trump would make a statement at 1AM, effectively giving the Republicans cover to vote for the president’s agenda. 1AM came and went, and Trump did not release a statement. There was nothing by 2AM. By 3AM, it was becoming clear that he wouldn’t weigh in at all. And at 4AM, with no president to hide behind, the rest of the Senate Republicans gave up.

The death of the AI moratorium marks a pivotal moment in the feud between the MAGA populists and the tech bros. Sure, Bannon and the rest of the populists have railed against the “broligarchy” on their podcasts since Trump’s victory in November, but the real action had been happening behind closed doors for months. The drama over Trump’s (dubiously legal) firings of Librarian of Congress and Register of Copyrights, for instance, started when the Copyright Office released a prepublication version of a report interpreting copyright law in a way that was somewhat unfavorable to AI companies. At first it seemed that the tech right, led by Elon Musk, had taken over the Copyright Office. Then it became clear that MAGAworld had opportunistically landed a blow against Silicon Valley by filling the positions with anti-Big Tech government lawyers. Elon Musk’s exile a month later was set in motion by populists inside the White House — who are ideologically aligned with Bannon and others — who’d convinced Trump that Musk’s people were disloyal. 

But this is the most meaningful and visible political win that the populists have notched in their feud with the tech industry. They would have won the battle if only four GOP senators had voted no. But they got 52 GOP senators voting no, including Cruz, who’d written the Senate’s moratorium in the first place — a blatant demonstration of MAGA political capital. 

Bannon was more than happy to take a victory lap and claim the win. “This is, I believe, our Lexington and Concord against AI,” he said afterward. “I’m not against AI. I’m against a completely unregulated AI driven by four people” — Sam Altman, Elon Musk, Dario Amodei, and David Sacks — “who do not have the best interests of our country or her citizens at the forefront of their mind.” (He added one more later: Demis Hassabis, the cofounder of Google DeepMind.) 

And in case it seems like Bannon and Davis might be too self-congratulatory, the interest groups, too, were quick to credit them the next morning. “It basically went down the way Mike Davis and Steve Bannon describe [it] on Bannon’s show,” said Jason Van Beek, the chief government affairs officer at the Future of Life Institute, a nonpartisan, nonprofit group dedicated to mitigating AI risks. He then sent The Verge a link to the latest episode of War Room.

Update, July 11th: The story has now been updated to include a response from Sen. Blackburn’s office.

Correction, July 11th: An earlier version of this article mischaracterized Mike Davis as a lawyer for Donald Trump. He does not personally represent the president.

For years, Meta trained its AI programs using the billions of public images uploaded by users onto Facebook and Instagram’s servers. Now, it’s also hoping to access the billions of images that users haven’t uploaded to those servers. Meta tells The Verge that it’s not currently training its AI models on those photos, but it would not answer our questions about whether it might do so in future, or what rights it will hold over your camera roll images.

On Friday, TechCrunch reported that Facebook users trying to post something on the Story feature have encountered pop-up messages asking if they’d like to opt into “cloud processing”, which would allow Facebook to “select media from your camera roll and upload it to our cloud on a regular basis”, to generate “ideas like collages, recaps, AI restyling or themes like birthdays or graduations.” 

By allowing this feature, the message continues, users are agreeing to Meta AI terms, which allows their AI to analyze “media and facial features” of those unpublished photos, as well as the date said photos were taken, and the presence of other people or objects in them. You further grant Meta the right to “retain and use” that personal information.

Meta recently acknowledged that it scraped the data from all the content that’s been published on Facebook and Instagram since 2007 to train its generative AI models. Though the company stated that it’s only used public posts uploaded from adult users over the age of 18, it has long been vague about exactly what “public” entails, as well as what counted as an “adult user” in 2007. 

Meta tells The Verge that, for now, it’s not training on your unpublished photos with this new feature. “[The Verge’s headline] implies we are currently training our AI models with these photos, which we aren’t. This test doesn’t use people’s photos to improve or train our AI models,” Meta public affairs manager Ryan Daniels tells The Verge.

Meta’s public stance is that the feature is “very early,” innocuous and entirely opt-in: “We’re exploring ways to make content sharing easier for people on Facebook by testing suggestions of ready-to-share and curated content from a person’s camera roll. These suggestions are opt-in only and only shown to you – unless you decide to share them – and can be turned off at any time. Camera roll media may be used to improve these suggestions, but are not used to improve AI models in this test,” reads a statement from Meta comms manager Maria Cubeta.

On its face, that might sound not altogether different from Google Photos, which similarly might suggest AI tweaks to your images after you opt into Google Gemini. But unlike Google, which explicitly states that it does not train generative AI models with personal data gleaned from Google Photos, Meta’s current AI usage terms, which have been in place since June 23, 2024, do not provide any clarity as to whether unpublished photos accessed through “cloud processing” are exempt from being used as training data — and Meta would not clear that up for us going forward.

And while Daniels and Cubeta tell The Verge that opting in only gives Meta permission to retrieve 30 days worth of your unpublished camera roll at a time, it appears that Meta is retaining some data longer than that. “Camera roll suggestions based on themes, such as pets, weddings and graduations, may include media that is older than 30 days,” Meta writes.

Thankfully, Facebook users do have an option to turn off camera roll cloud processing in their settings, which, once activated, will also start removing unpublished photos from the cloud after 30 days.

The feature suggests a new incursion into our previously private data, one that bypasses the point of friction known as conscientiously deciding to post a photo for public consumption.  And according to Reddit posts found by TechCrunch, Meta’s already offering AI restyling suggestions on previously-uploaded photos, even if users hadn’t been aware of the feature: one user reported that Facebook had Studio Ghiblified her wedding photos without her knowledge.

Correction, June 27th: An earlier version of this story implied Meta was already training AI on these photos, but Meta now states that the current test does not yet do so. Also added statement and additional details from Meta.

Last year, Justin Sun was facing a federal investigation into his crypto empire, to the point that he avoided entering the US. Now, his company is going to be publicly listed on the American stock markets – and reportedly with the assistance of Eric Trump.

Tron, Sun’s digital assets platform, announced on Monday that it plans to go public in the U.S. via a reverse merger with SRM Entertainment, a Nasdaq-listed company that previously designed merchandise for theme parks before pivoting to crypto. At the same time, Trump’s sons announced the launch of Trump Mobile, an MVNO that will also sell a $499 gold Android phone. 

On the company’s site, SRM states that it has worked with Walt Disney World, SeaWorld, and Sesame Place. SRM says it has raised $100 million from a private investor to buy the Tron tokens (TRX), with plans to issue shares and warrants valuing the deal at up to $210 million.

The deal was brokered by Dominari Securities, a New York-based boutique investment bank operating out of Trump Tower, whose board includes both Donald Trump Jr. and Eric Trump, the president’s sons. Shortly after Trump’s election, Sun became an advisor and investor in World Liberty Financial, the Trump family’s crypto company, and promptly bought $75 million of their memecoin, $TRUMP, when the token launched in January. 

By the end of February, the SEC had officially paused its investigation into Sun’s businesses. Sun continued to invest in Trumpworld entities, both in crypto and elsewhere: in May, he won a private dinner with the president by buying $16 million worth of $TRUMP in a contest.

According to an unnamed source cited by the Financial Times, Eric Trump, who oversees the Trump Organization’s private sector holdings, is expected to take a position at the company, which will be renamed Tron, Inc.