Anthropic heads to Washington to win its AI models back | Official Website of Louis Velazquez, entrepreneur, finance guy and tech innovator

Markets like to believe that size is a kind of armor. The bigger a company gets, the safer it looks, until it stops feeling like a business and starts feeling like part of the furniture of the economy itself.

That belief is doing a lot of work right now. Investors have spent the past two years treating the leaders of the artificial intelligence (AI) boom as permanent fixtures, the load-bearing walls under the S&P 500 and under a generation of retirement accounts. Their products feel less like services you could lose and more like infrastructure, as fixed as the power grid. The assumption underneath is simple. Companies worth this much do not just stop.

Then one of them did.

On Friday, June 12, the U.S. government effectively switched off the flagship product of a company valued at close to a trillion dollars. It did not take a hack, a crash, or a courtroom. It took a letter.

The company is Anthropic, maker of the Claude AI assistant, and by Monday, June 15, its senior staff were in Washington trying to turn its most powerful models back on.

How a single letter pulled Anthropic offline

The order came from the Commerce Department. Secretary Howard Lutnick sent Anthropic chief executive Dario Amodei an export-control directive applying to its newest models, Fable 5 and Mythos 5, barring their use “by any foreign national,” according to CNBC.

The directive covered any location outside the U.S. and every foreign national inside it, “including Anthropic’s own non-citizen employees,” according to Axios. Because the company could not realistically wall those users off, it pulled both models for everyone.

More on AI

When I traced June 12’s timeline, the speed was the part that stood out. The White House called Anthropic that afternoon, said the models posed a national security threat, and gave the company 90 minutes to disable them, Axios confirmed. The letter arrived hours later, requiring a federal license for any export or transfer of the models and warning of civil penalties.

Anthropic disputed the basis for the shutdown. The company called the directive a likely misunderstanding and said it was working to restore access as soon as possible, pushing back on the idea that a narrow vulnerability should justify recalling a product already in customers’ hands, TheStreet noted.

Anthropic’s most powerful AI went dark. Now it’s in D.C.

NurPhoto / Getty Images

What Fable 5, Mythos 5 shutdowns reveal about AI regulatory risk

Here is the part that should reach past the AI crowd and into ordinary portfolios. A company most Americans cannot buy a single share of just had its best product turned off by the government over a weekend, and the value at stake is enormous.

You may not own Anthropic, but you almost certainly own its weather. The same export-control regime that froze Fable 5 governs the chips that Nvidia (NVDA) and AMD (AMD) sell abroad, and a generation of index funds now leans hard on a handful of AI names.

When Washington shows it can disable a frontier model on a Friday night, that is a new line item of risk for every AI-exposed fund, even the ones tucked inside a 401(k).

The trigger, by the company’s account, was research showing a way to “jailbreak” Fable 5, or coax it past its safeguards. Those findings were produced by Amazon (AMZN) researchers, an awkward detail, given that Amazon is one of Anthropic’s largest backers and its main cloud host.

Independent security voices were not impressed. “If national defense is the goal, this is an own goal,” said Katie Moussouris, the Luta Security chief executive who built Microsoft’s (MSFT) first bug-bounty program, in comments to Fortune.

She described the research as defensive rather than offensive, a tool meant to help cyber defenders, not attackers.

Why Washington and Anthropic keep colliding

This is not a sudden feud. It is the latest round of a fight that has been escalating since winter, and the pattern matters more than any single letter.

The flashpoints, in order:

In February 2026, the Pentagon declared Anthropic a supply-chain risk to national security, a label usually reserved for foreign firms like Huawei, according to TheStreet.
President Donald Trump then directed federal agencies to stop using Anthropic technology and publicly called it a “radical left, woke company,” according to that same reporting.
The clash traced back to Anthropic’s refusal to let its models be used for autonomous weapons or mass surveillance of U.S. citizens, the red lines the company wrote into its own policies, as highlighted in my TheStreet report.
On Tuesday, June 9, Anthropic released Fable 5 and Mythos 5 to the public after working with agencies on testing, according to CNBC.
Three days later, on June 12, the Commerce letter pulled them back, according to Axios.

What strikes me, reading Anthropic’s statements across these episodes, is how consistent the company has been. Amodei has cast the standoff in patriotic terms, calling his team “patriotic Americans” acting for U.S. national security, even while refusing the military and surveillance uses the administration wanted.

The government’s case is narrower and blunter. Officials treat the most capable models as weapons-grade technology, subject to the same export rules as advanced chips, and they were unsettled by a system that could probe cyber defenses on its own.

Where the standoff leaves AI investors

The timing carries real money. Anthropic closed a $65 billion funding round at a $965 billion valuation last month, a figure that briefly eclipsed OpenAI, according to TechCrunch. It then confidentially filed for an initial public offering (IPO) that could land this fall.

An IPO at that scale would push Anthropic shares into the same funds, pensions, and brokerage accounts that already ride the AI trade through Nvidia and Microsoft.

The question every prospective holder now has to weigh is one that did not exist a week ago. What is a frontier AI company worth if a single agency can switch off its crown jewel between dinner and bedtime?

For now, Anthropic’s technical team is meeting White House officials in person and has spoken with them daily since the directive landed, according to The Hill. Both sides have reason to find a face-saving fix, and the company says it expects to restore access.

The deeper signal will outlast this particular truce. Washington has decided that the line between a commercial AI product and a national security asset is one it gets to draw, and redraw, in real time.

Anyone betting their savings on the AI boom should learn to read that line as carefully as they read an earnings report.