Just now, a news story flooded the headlines:

The U.S. government has directly placed export controls on Anthropic’s two strongest models, Fable 5 and Mythos 5.
I just tried with my Australian account, and indeed, it doesn’t work.
In the past few days, I’ve actually tested Fable 5 quite a bit; if you don’t use it for multi-tasking and just for assisting research and investment, it’s very good—expensive for a reason.
Why did this happen?
Commerce Secretary Howard Lutnick sent a letter to Dario Amodei on Friday, listing Mythos 5 and Fable 5 under export controls, prohibiting access from anywhere outside the U.S. and to all foreigners within the country.
The trigger was another company claiming to have successfully jailbroken Mythos, which raised government alertness about national security risks.
A more critical background is that the government previously tried to get Anthropic to delay the release of these two models but failed, so they directly used the export control card.
The official documents mention specific concerns that the models could bypass cybersecurity defenses through certain prompts.
Anthropic strongly disagrees, claiming Fable’s safety barriers have been tested for thousands of hours by red teams, surpassing any deployed models, and that no “general jailbreak” has ever been found.
An interesting paradox:
Anthropic is now simultaneously on the Pentagon’s blacklist (considered too dangerous for government use) and under the Department of Commerce’s licensing system (considered too dangerous for foreign use).
Essentially, Washington is treating cutting-edge AI as a military-grade strategic asset.
What about foreign employees?
This is the harshest part of the directive. The ban explicitly covers “any foreigners, whether inside or outside the U.S., including Anthropic’s foreign employees.”
This means Anthropic’s non-American engineers cannot access these two models.
To ensure compliance, Anthropic has simply disabled Fable 5 and Mythos 5 for all customers, but all other Claude models remain unaffected.
In practice, this is a “one-size-fits-all” approach—because it’s impossible to precisely distinguish user nationality at the product level.
So, currently, all customers, regardless of whether they are in the U.S. or not, can’t use 😂😂😂.
And Americans are also getting annoyed.
How long will this last?
Likely, it’s short-term and not permanent.
A government official told Axios that the models need to be “locked down” within the U.S. government’s national security system for “reinforcement,” and this could happen within the next few weeks.
Anthropic has also characterized it as a “misunderstanding,” saying they are working to restore access as soon as possible.
So, this appears more like regulatory pressure and negotiations rather than a final verdict.
However, even if access is restored in the future, the compliance framework might remain long-term, as these models are already being treated as strategic materials. 🥲
What about the impact on IPO?
The timing is very delicate right now.
Anthropic just secretly filed an S-1 on June 1, with an estimated valuation of about $965 billion, aiming to go public in October, led by Goldman Sachs and J.P. Morgan.
My assessment has two layers:
In the short term, it’s negative noise, but timing is everything.
Anthropic just filed the S-1 on June 1, aiming for a October IPO with nearly a trillion-dollar valuation—this is the first case testing whether “cutting-edge AI valuation can withstand the public market.”
The risk section of the prospectus now needs an additional line: the government can shut down your flagship product at any time.
Underwriters’ pricing will likely incorporate a heavier regulatory discount.
But conversely, the more capable the models are, the more they attract government scrutiny—this news, to some extent, is a testament to Mythos’s strength.
If within a few weeks, the ban is lifted and Anthropic spins it as “we’re too advanced, so we’re being regulated,” it might not be a net negative for long-term valuation.
The real question is: is this a one-time event, or a sign of ongoing deterioration in the relationship between the government and Anthropic after being blacklisted by the Pentagon?
If it’s the latter, “policy risk” could become a structural valuation suppression factor relative to OpenAI.
Let’s see if this crisis might create a good entry point for pre-IPO investments.