This is an edition of The Atlantic Daily, a newsletter that guides you through the biggest stories of the day, helps you discover new ideas, and recommends the best in culture. Sign up for it here.

Earlier this week, Secretary of Defense Pete Hegseth sat down with Dario Amodei, the CEO of the leading AI firm Anthropic, for a conversation about ethics. The Pentagon had been using the company’s flagship product, Claude, for months as part of a $200 million contract—the AI had even reportedly played a role in the January mission to capture Venezuelan President Nicolás Maduro—but Hegseth wasn’t satisfied. There were certain things Claude just wouldn’t do.

That’s because Anthropic had instilled in it certain restrictions. The Pentagon’s version of Claude could not be used to facilitate the mass surveillance of Americans, nor could it be used in fully autonomous weaponry—situations where computers, rather than humans, make the final decision about whom to kill. According to a source familiar with this week’s meeting, Hegseth made clear that if Anthropic did not eliminate those two guardrails by Friday afternoon, two things could happen: The Department of Defense could use the Defense Production Act, a Cold War–era law, to essentially commandeer a more permissive iteration of the AI, or it could label Anthropic a “supply-chain risk,” meaning that anyone doing business with the U.S. military would be forbidden from associating with the company. (This penalty is typically reserved for foreign firms such as China’s Huawei and ZTE.)

This evening, Anthropic said in a public statement that it “cannot in good conscience accede” to the Pentagon’s request. What happens next could mark a crucial moment for the company, and for the American government’s approach to AI regulation more broadly. In refusing to bow to an administration that has been intent on bullying private companies into submission, Amodei and his team are taking a bold stand on ethical grounds, and risking a censure that could erode Anthropic’s long-term viability.

During the first year of Donald Trump’s second term, the White House had a more relaxed attitude toward AI regulation; an AI Action Plan from July stresses that the administration will “continue to reject radical climate dogma and bureaucratic red tape” to encourage innovation. Hegseth is now, in effect, threatening to partially nationalize one of the biggest AI players in the private sector—and force the company to go against its own principles. “This is the most aggressive AI regulatory move I have ever seen, by any government anywhere in the world,” Dean Ball, who helped write some of the Trump administration’s AI policies, told me.

The Pentagon has already reportedly been reaching out to other defense contractors to see if they’re connected to Anthropic, a sign that officials are preparing to designate the company a supply-chain risk. Now that Anthropic has defied Hegseth, the contract is likely in peril. The firm doesn’t really need the $200 million—it reportedly pulls in $14 billion a year, and it said it raised $30 billion in venture capital just weeks ago—but being blacklisted could affect its ability to scale up in the future. (“We are not walking away from negotiations,” an Anthropic spokesperson told The Atlantic in a statement. “We continue to engage in good faith with the Department on a way forward.” The Pentagon told CBS on Tuesday that “this has nothing to do with mass surveillance and autonomous weapons being used,” and that ”the Pentagon has only given out lawful orders.”)

As AI firms around the world jockey for dominance, Anthropic has distinguished itself by emphasizing safety. OpenAI’s ChatGPT has been criticized for playing up some users’ delusions, leading to cases of “AI psychosis,” and just last month, xAI’s Grok was spinning up nearly nude images of almost anyone without consent. (xAI has said it is restricting Grok from generating these kinds of images, and OpenAI has said it is working to make ChatGPT better support people in distress.) Meanwhile, Anthropic’s consumer-facing chatbot doesn’t generate images at all. By refusing to cave to government pressure, it may have just averted another crisis: a major public backlash from consumers, some of whom see the company as a more principled player in the AI wars. Anthropic recently faced some pushback over changing its policies—Time reported on Tuesday that, in a seemingly unrelated move, the company dropped a core safety pledge concerning its broader approach to AI development.

Weeks before Hegseth issued his ultimatum, Amodei opined on his website about the risks involved with precisely the two guardrails the Pentagon is targeting. “In some cases,” he wrote, “large-scale surveillance with powerful AI, mass propaganda with powerful AI, and certain types of offensive uses of fully autonomous weapons should be considered crimes against humanity.”

The Trump administration doesn’t seem to know what it wants from AI. On one hand, it’s deeply suspicious of certain kinds of models. The White House’s designated AI czar, David Sacks, has criticized Anthropic for “running a sophisticated regulatory capture strategy based on fear-mongering,” essentially accusing the firm of pushing for unnecessary, innovation-squashing limitations and jeopardizing the future of American tech. The administration has also criticized AI bots for sometimes spitting out “woke” replies. On the other hand, Claude is apparently valuable enough that it’s on the cusp of being commandeered by the federal government.

Ball told me that the Department of Defense may have a point—that there’s an argument to be made about reining in Silicon Valley’s control over the government’s use of new technologies. Although the concentration of power among the technocratic elite is certainly troubling, Hegseth’s proposed punishments for Anthropic are misguided and plainly contradictory. The Defense Production Act does allow the government to intervene in domestic industries in the interest of national security (the Biden administration invoked it in a 2023 executive order on AI regulation). But is Claude so important for U.S. national security that the government needs to compel Anthropic to create an untethered new version? Or is it so dangerous that it needs to be shunned—not just by the Pentagon, but by any business connected to the military? A third, even-more-bewildering option is also on the table: Hegseth could decide to simultaneously commission a modified Claude and sanction the company that stewards it.

All of this ignores a much simpler solution: Hegseth could just start a partnership with a different firm. It’s a good time for his department to be in business with tech, since the mood of Silicon Valley has lately become much more Pentagon-friendly. Palantir’s Alex Karp has touted that his software is used “to scare our enemies and, on occasion, kill them”; the technologist and entrepreneur Palmer Luckey is already building autonomous weaponry for the government; and Andreessen Horowitz’s American Dynamism funds are helping funnel the country’s top young minds into defense tech. But rather than look elsewhere, Hegseth is threatening to crush Anthropic—implying that if he can’t control Claude, no one can.

As the defense secretary looks to make an example of the company, he’s taking a cue from Trump, who has used legal and extralegal pressure to effectively force other private businesses, particularly big law firms, banks, and universities, into submission. These acts of coercion have the potential to reshape American capitalism: We are beginning to see a market where winners and losers are decided less by the quality of their products and more by their seeming fealty to the White House. How that will affect the success of businesses and the economy is uncertain.

The Pentagon created this ultimatum precisely because it understands Anthropic’s world-altering potential. The administration just can’t decide if it’s an asset, a liability, or both.

Related:

• The high-stakes fight between Hegseth and Anthropic
• Anthropic is at war with itself.

Here are three new stories from The Atlantic:

• The first couple of a dysfunctional DHS
• Trump’s favorite voter-ID bill would probably backfire.
• Adam Serwer: How America chose not to hold the powerful to account

Today’s News

1. A Columbia University student detained this morning by federal immigration agents has been released. The arresting officers reportedly misrepresented themselves as looking for a missing child in order to gain entry to the student’s residential building.
2. Hillary Clinton told the House Oversight Committee that she has no new information about Jeffrey Epstein and maintained that she had no knowledge of his crimes; she criticized congressional Republicans’ handling of the probe as partisan. Bill Clinton is scheduled to give his deposition tomorrow.
3. Cuban forces killed four people and wounded six after firing on a Florida-registered speedboat that Cuban authorities say entered the country’s waters yesterday and opened fire on a patrol vessel. Cuba claims that the U.S.-based passengers were armed and planning a “terrorist” infiltration.

Dispatches

• Time-Travel Thursdays: Misbehaving dogs in the White House have plagued administration after administration, Nicholas Florko writes.

Explore all of our newsletters here.

More From The Atlantic

• Eliot A. Cohen: The degraded State of the Union
• The tariffs loss is paradoxically a win for Trump.
• Well, that didn’t sound like Casey Means.
• Enough with the bros.
• The Democrats who got weird during the State of the Union
• Radio Atlantic: What the Pentagon fears in Iran

Evening Read

This Looks Like an Insider Bet on Aliens

By Ross Andersen

On Monday night, someone placed a peculiar bet on the prediction market Kalshi. At 7:45 p.m. eastern time, a single trader put down nearly $100,000 on the claim that, by the end of December, the Trump administration will confirm that alien life or technology exists elsewhere in our universe. According to The Atlantic’s review of Kalshi’s trading data, about 35 minutes after this bet was executed, it was followed by another that was almost twice as large (possibly from the same person). These were market-moving events: For one brief stretch, the market appeared to think that there was at least a one-in-three chance that the U.S. government will announce the existence of aliens this year. Perhaps this was just some overexcited UFO diehard with a hunch and money to burn. Or maybe, as some observers quickly noted, it was a trader with inside knowledge.

Read the full article.

Culture Break

Explore. When did literature get less dirty? A puritan strain is manifesting in realist novels as a marked absence of straight sex, Lily Meyer writes.

Read. Casey Schwartz on two new books that demonstrate how Martha Gellhorn, Janet Flanner, and other female reporters took journalism in directions that men could not.

Play our daily crossword.

Rafaela Jinich contributed to this newsletter.

When you buy a book using a link in this newsletter, we receive a commission. Thank you for supporting The Atlantic.

As an aging man, one has experiences. However, there are some things that man should not know and one of them is three feet of snow in front of your house.

I went to bed last night about 11. It was snowing a little bit. As of 1 PM, the Providence airport (and I live about 2 miles from there so it’s always going to be very close to accurate for me) had 32.8 inches of snow. The old record, in the infamous Blizzard of 78, was 28.6 inches. That’s just shattering the record. We’ve definitely had a couple of inches since then so the final total is going to be right around 3 feet. It’s a hell of a thing.

I guess I’ve adjusted to being a New Englander because my wife and I went out and dug the whole thing out already too.

Anyway, hope your snow experiences have been less extreme!

The post Snowocalypse appeared first on Lawyers, Guns & Money.

Pirates have it rough. In order to board a cargo ship, they need to get within range using a pirate "mothership," typically a hijacked fishing vessel. Once close enough, the mothership launches several small, high-powered skiffs. These skiffs close the distance with the cargo ships, typically traveling through the cargo ship's tumultuous wake to avoid detection. Once alongside the ship, the pirates then have to use grappling hooks or ladders to climb to the deck; not so easy when you've got an AK-47 slung across your back and the cargo ship's crew is desperately trying to douse you with fire hoses.

That's the old Somali model, popular around a decade ago. But current-day Houthi rebels have used technology to change the game. Because they're driven by ideology and not profit, the Houthis don't need to board the ships; they just need to mess them up. They've thus been early adopters of Uncrewed Surface Vehicles (USVs), essentially kamikaze robot boats packed with explosives. These are relatively easy to build—they're just regular skiffs rigged up with GPS and remote control. It's a lot cheaper than building your own navy.

Worse, the Houthis have been getting friendly with Somali pirates. And so, DARPA reckons, it's just a matter of time before Somali pirates start using the Houthi model. Rather than risking life and limb scaling the side of a ship, all they need to do is show up with a bomb-laden USV. They contact the crew of the cargo ship, and explain that unless a $XXX,XXX ransom is delivered, they'll blow the ship up.

DARPA has pre-emptively developed a concept to defend against this: The Pulling Guard. This is a small, unmanned, armed platform that cargo ships would tow behind them. In the event of trouble, the Pulling Guard autonomously sends up a quadrotor drone to get the big picture. A remote operator then views the feed, assesses the threat, and can opt to fire missiles from the Pulling Guard.

DARPA points out that all of the technology to create a Pulling Guard already exists; their focus, then, is on "marinizing" the sensors and systems, and packaging them in a modular way in order to ease manufacture. If they can pull it off, the Houthis will have to go back to the drawing board.

Congratulations again to all the marginal voters who put Trump in office because prices were too high:

President Donald Trump said Saturday that he wants a global tariff of 15%, up from 10% he had announced a day earlier after the U.S. Supreme Court struck down many of the far-reaching taxes on imports that he had imposed over the last year.

Trump’s announcement on social media was the latest sign that despite the court’s check on his powers, the Republican president still intends to ratchet up tariffs in an unpredictable way. Tariffs have been his favorite tool for rewriting the rules of global commerce and applying international pressure.

The court’s decision on Friday struck down tariffs that Trump had imposed on nearly every country using an emergency powers law. Trump now said he will use a different, albeit more limited, legal authority.

He’s already signed an executive order enabling him to bypass Congress and impose a 10% tax on imports from around the world, starting on Tuesday, the same day as his State of the Union speech. However, those tariffs are limited to 150 days unless they are extended legislatively.

Timing this with the State of the Union so these tax increases* get maximum attention makes sense if spite is the primary mover of your governance.

*Although don’t worry: according to Clarence Thomas’s latest royalist opinion, tariffs aren’t taxes, merely extractions one has to pay the government:

The post The taxes will increase until morale improves appeared first on Lawyers, Guns & Money.

I really really hate the entire world of “election reforms” to “give people more choices.” They are, to me at least, a classic example of “let’s get the politics out of politics” moves that end up serving special interests, along the lines of term limits, as one egregious examples. They also suffer from not recognizing that the average American voter is a moron and so pretend to believe that if you just give people more choices, the people will consider all the options and make the right choice for them and that’s democracy and is good.

Instead, you risk what actually could happen in California if all these big egos don’t get out of their own way, which is the foolishness of the jungle primary leading to this very blue state having a final choice between two Republicans for governor this fall.

And yet, despite this huge partisan tilt, there’s a very real chance that the state will elect a MAGA Republican governor this November. Not that the two Republicans seeking that office are in any way popular: The RealClearPolitics polling average shows one favored by only 15 percent of voters, and the other by 13 percent. But every one of the eight Democrats also seeking the office is polling lower than that in the most recent surveys.

The culprit here is the state’s absurd jungle primary, a measure California adopted in 2010. Partisan primaries in the state have been condensed into a single June primary in which candidates of all parties (or no party) appear on the same ballot, with the top two proceeding to a November general election, where no write-in votes for other candidates are permitted.

The reason for that switch is that in 2009, state budgets required two-thirds majorities in each house of the legislature (they now require just a simple majority), and the Democrats—not yet commanding the level of support they’ve secured since—were one vote shy of that total in the Senate. They needed the vote of Abel Maldonado, the one moderate Republican in that body. But Maldonado, who was eyeing a future gubernatorial run, demanded they put a measure on the 2010 ballot that would scrap party primaries for the jungle. Maldonado and the state’s moderate Republican governor at the time, Arnold Schwarzenegger, calculated that this would lead to more moderate elected officials, though in the years since every moderate Republican in the state, including Schwarzenegger, has been driven from the party’s ranks.

Still, the idea sounded unobjectionable to voters at that time, and only a handful of pundits opposed it. I was one of those opponents, writing here and in the Los Angeles Times that the jungle might condemn the state to an elected leadership that’s hugely out of sync with state voters, if only two members of one party ran for an office that a passel of members of the other party were also seeking. After all, as California is an overwhelmingly Democratic state, more Democrats invariably run for statewide office than Republicans do.

That’s Harold Meyerson writing this and he’s right about almost everything, so of course he was right about this too. Fundamentally, the problem is ego, the great problem of all politicians:

In fact, the field is still in flux. For a long time, the leading candidate has been former Rep. Katie Porter, who won a reputation as a feisty progressive during her tenure in Congress. But Porter has been out of Congress, and largely out of the news, for several years now, and had some rocky personal appearances during her gubernatorial rollout. In the most recent polls, she’s running slightly behind the two Republicans, and in a virtual tie with Bay Area Rep. Eric Swalwell, a frequent cable news guest who is little known in Southern California.

Two former elected officials lag behind them: onetime Los Angeles Mayor Antonio Villaraigosa, out of office for the past 13 years, during which his politics have moved decidedly rightward; and Xavier Becerra, who was Joe Biden’s secretary of health and human services and, before that, California’s attorney general. Two other statewide elected officials are lagging even behind them, reinforcing the notion that the down-ticket statewide offices guarantee almost total obscurity: former state Controller Betty Yee and current state Superintendent of Public Instruction Tony Thurmond, each of whom commands a level of support in the low single digits. Yee’s almost universally unknown record situates her, along with Porter, in the party’s progressive wing. Along with Thurmond—and none of the other candidates—she supports the wealth tax on the state’s billionaires that may come before voters on November’s ballot if it collects the requisite number of signatures.

Every big California political player, starting with the former Mr. Kimberly Guilfoyle in the governor’s mansion, needs to sit down with everyone but Swalwell and Porter and do whatever is necessary for them to drop out. Villaraigosa is so 20 years ago, Becerra is hopeless, the others have just enough support to help elect a Republican. Honestly, Porter has become ridiculous herself. Policywise, she’d be great, but she’s actively mean to a lot of her own staff and that has spilled into her relationships with journalists. Whatever it takes, clear all but two people out of here. Any Democrat is going to win this race. But one of them has to be in the top two.

“Good government liberals” love things like electoral reforms of all stripes and they are both a waste of time and counterproductive. Kill the jungle primary everywhere.

The post The Impending California Disaster appeared first on Lawyers, Guns & Money.

Manhattan started extending its land mass back in the colonial era, using construction debris, sunken ships, ashes, ballast, and other waste to reclaim land and enlarge the island.

Pearl Street used to be the southern boundary; Greenwich Street was at the edge of the Hudson River. Manhattan continued to grow in the 19th century, but by the early 1900s—with almost all of Manhattan urbanized—civil engineers were considering new ways to create more real estate.

Enter a highly esteemed and successful engineer of bridges and skyscrapers named T. Kennard Thomson.

His proposal, popularized in a nationally syndicated newspaper article in June 1911, was to extend Manhattan four miles into New York Harbor, adding 4,100 acres to New York City’s most populated borough.

“The method of reclamation to be followed is extremely simple,” he told a reporter. “I would merely erect concrete seawalls from the Battery toward Staten Island for the desired length, and then fill them in.”

Thomson made his case by focusing on the taxes that could be collected on the additional land. But he was especially concerned with the journey ships took from the harbor to the docks in Manhattan.

Extending the island into the harbor while preserving a narrower ship canal above Staten Island would make it easier for ships to complete their voyage, he believed. The more ships that dock in Manhattan, the more enriched city coffers become.

Part of his proposal involved building a “six-track subway all around Manhattan Island, including the new extension. The subway would be built underneath the present dock line of the city.”

Visionary or pipe dreamer, Thomson was grounded enough to know that he needed city officials to get on board with his plan. The article states that his proposal was “under consideration, and other engineers who have looked into the matter regard it as entirely feasible.”

You can imagine what City Hall must have thought of this massive, likely quite expensive plan. But Thomson wasn’t finished coming up with new ideas for enlarging Manhattan.

In 1916, he published a proposal in Popular Science not only to build into New York Harbor but to fill in the entire East River (third image), which would reclaim 50 square miles and create “a really greater New York.”

You know the end of this story. Like so many other fantastical ideas that never came to pass, Thomson’s second proposal never came to fruition.

[Top image: The Atlanta Journal; Second image: Geographicus.com; third image: Popular Science; Fourth image: New York Tribune]