Anthropic’s AI models were sidelined by the White House over potential security risks, triggering a scramble that includes a months-long safety debate, a dispute with the Department of Defense over “supply chain” risk, and an export ban that led the company to
For Dario Amodei, the past year has been a sprint with no safe finishing line.
On Friday, the newest shock landed from Washington: the White House sidelined Anthropic’s new models over potential security risks. The move came as the company was already navigating a high-wire stretch between competing in the AI race and arguing—loudly—for safety rules that keep society protected.
The push-and-pull didn’t start with the ban. It had been building for months in public, and then tightening in ways that reached far beyond the AI lab.
Back on Jan. 27, Anthropic CEO Dario Amodei released a 19,000-word message on the future of AI and the “serious civilizational challenge” it poses. Less than a month later, on Feb. 24, Anthropic weakened its foundational safety commitment amid heightened competition and a lack of government regulation.
By Feb. 27, the stakes turned sharper inside the national security world. A dispute between Anthropic and the Department of Defense over how its AI models are used led the DoD to label Anthropic a supply chain risk.
Anthropic kept trying to steer the conversation toward safety—but also pushed the boundary of what its systems could do. On April 7, the company said its Mythos model was too powerful for the public, citing its knack for finding “high-severity vulnerabilities.”
Then came the company’s move toward Wall Street. On June 1, Anthropic filed a confidential S-1 for its IPO.
That same week, Anthropic argued for restraint across the frontier AI race. On June 5, it called for a coordinated slowdown among frontier AI labs “to enable societal structures and alignment research to keep up with the advance of the technology.”
Amodei’s own words followed the contradiction of speed versus control. On June 10, he published a blog saying AI is moving at a “lightning pace” while policy is “moving very slowly.”
Two days later, enforcement arrived. On June 12, a US order barring foreign entities or individuals from using Anthropic’s Fable 5 and Mythos led the company to disable the models for everyone. The company’s safeguards narrative collided with the blunt reality of export rules.
By June 15, the pressure turned into negotiation: Trump officials reportedly met with Anthropic to resolve the export ban on Fable.
And amid that sequence, Anthropic released what it said was a safer version. On June 9, it released Fable 5, built by putting safeguards on the Mythos model.
The story, even at a distance, lands as a pattern of competing imperatives: those best positioned to warn about advanced AI dangers are also the ones building the systems governments now feel compelled to restrict.
The timeline shows the same tension repeatedly tightening—safety commitments tested by competition, national security concerns surfacing as use disputes, and then sweeping restrictions that force changes across a whole model lineup.
With the White House sidelining the new models over potential security risks. the immediate consequence is clear: Anthropic’s next step is no longer only about engineering or product strategy. It is about who has the authority to decide what counts as “safe enough. ” and how quickly that definition can catch up with the pace the industry itself is accelerating.
Anthropic Dario Amodei White House AI safety export ban Mythos Fable 5 Department of Defense S-1 IPO frontier AI slowdown
So they banned it cause it’s “too powerful”? Sounds like fear-mongering.
Wait I thought the whole point was security, but now it’s like an export ban and a supply chain thing? Who even decides what counts as a risk. This feels like politics more than AI safety.
The DoD called it a “supply chain” risk which makes me think they mean like the chips or servers or whatever. So are we blaming the hardware manufacturers now? Also the IPO filing is probably why they got side-eyed, like money first then safety later.
Honestly this whole “safety debate” just sounds like everyone arguing over who gets access first. The article keeps mentioning a bunch of dates like it’s a timeline for doom. First export ban, then “too powerful” Mythos, then sidelined again… so is it blocked or not blocked? And if it’s “for security,” why keep letting them publish half the stuff.