All right. So, we need to talk about
what just went down with Anthropic and
their Fable 5 model because this is
honestly one of the wildest things I've
seen happen in the AI space in a while.
So, on Friday, June 12th, 2026, at
exactly 5:21 p.m. Eastern Time,
Anthropic received a letter from the US
government. And this wasn't just any
letter. This was basically the
government saying, "Shut it down. Now."
They had to disable access to both Fable
5 and Mythos 5 for every single one of
their customers. Not just international
users, not just specific countries,
literally everyone. And the reason?
The government cited national security
concerns under export control
authorities.
The shocking part was that the directive
said Anthropic had to suspend access to
Fable 5 and Mythos 5 for every foreign
national, whether they were inside or
outside the United States.
And get this, that includes Anthropic's
own foreign national employees. So,
you've got people who literally built
these models, and now they're not
allowed to access their own work.
>> [music]
>> That's how extreme this order was.
Anthropic had no choice but to pull the
plug on both models completely to ensure
compliance. But all their other models,
including regular Claude, are still
working fine. So, let me give you some
context here because this didn't just
come out of nowhere. Anthropic had just
launched Fable 5 on June 9th, literally
3 days before this happened. Fable was
designed to bring many of the
capabilities of Mythos to the public.
Now, Mythos is their state-of-the-art
cybersecurity model that was previously
only available to a very select group of
companies and research partners through
something called Project Glasswing.
We're talking about a model so advanced
that during testing, Fable 5 actually
beat Pokémon Fire Red, while the regular
Claude model had failed to [music] beat
the original Pokémon Red that it was
based on. That gives you an idea of how
much more capable this thing was.
The government's letter didn't actually
spell out the specific national security
concerns, which is already pretty
sketchy if you ask me. But Anthropic's
understanding is that someone
demonstrated a jailbreak method for
Fable 5 to government officials, and
that freaked them out enough to issue
this unprecedented order. According to
Anthropic, they reviewed the
demonstration, and it showed the
jailbreak being used to identify what
they describe as a small number of
previously known minor vulnerabilities.
And here's the kicker. Anthropic says
these vulnerabilities are pretty simple,
and that other publicly available
models, including OpenAI's GPT-5.5,
can discover the same things without
even needing a jailbreak. Now, let's
talk about what Anthropic did to try to
prevent this kind of thing in the first
place. [music] They've been super vocal
about the safeguards they put in place
for Fable. They worked with the US
government, the UK AISI, multiple
private third-party organizations, and
their own internal teams to red team
Fable's safeguards for literally
thousands of hours before launch. They
instituted really strong protections to
reduce the likelihood that Fable could
be misused for cybersecurity-related
tasks, and apparently these safeguards
were so aggressive that many users were
actually complaining they were too broad
and getting in the way of legitimate use
cases. Anthropic's whole approach was
what they called defense in depth. They
knew that perfect jailbreak resistance
isn't possible. Literally no model
provider can guarantee that. So, their
strategy was to make jailbreaks either
very narrow, meaning they only work in
specific circumstances, or extremely
expensive to produce if someone wanted a
universal jailbreak that could bypass
the model's safeguards across the board.
[music]
They combined this with thorough
monitoring to quickly detect and shut
down any successful attacks. This is
actually why they required 30-day
retention of customer data with Fable,
which was a policy change that upset
some privacy-focused users,
but Anthropic said they needed it to
research and mitigate [music] jailbreaks
effectively. In their statement,
Anthropic makes it pretty clear they're
not happy about this. They say they
believe the government should have the
ability to block unsafe AI deployments,
but it should be part of a statutory
process that's transparent, fair, clear,
and grounded in technical facts.
And this action, [music] according to
them, doesn't meet any of those
criteria. They point out that they
received only verbal evidence of this
potential jailbreak, and the technique
basically just involves asking the model
to read a specific code base and fix
[music] software flaws. Anthropic argues
that this capability is widely available
from other models and is literally used
every day by security defenders who are
trying to keep systems safe. Now, this
isn't Anthropic's first clash with the
Trump administration. Earlier in 2026,
the Department of Defense actually
labeled Anthropic as a supply chain
risk, which is a designation that's
historically been reserved for foreign
adversaries. This designation means
defense contractors have to certify they
won't use Claude models in their work
with the military. The whole thing
happened after negotiations between
Anthropic and the DoD [music] collapsed,
reportedly because Anthropic wouldn't
agree to let the military use their
models for certain purposes.
Anthropic actually sued the Trump
administration over this blacklisting,
and that litigation is still ongoing.
The reaction from the tech community has
been all over the place. Dean W Ball
from the Foundation for American
Innovation called it just baffling,
pointing out the contradiction of an
administration that's okay with
exporting advanced AI chips to China,
but wants to ban every non-American on
Earth from using our best models.
>> [music]
>> Marc Andreessen posted about it, clearly
not thrilled with the situation. Peter
Gurnis made this really interesting
observation that Anthropic kind of wrote
their own legal predicament by marketing
Mythos as too dangerous to release.
He said that if you describe your
product as a munition in every press
release, eventually a government is
going to take you at your word. Chris
Maguire from the Council on Foreign
Relations said that while he thinks
targeted export controls on Model Access
can be prudent, imposing
across-the-board controls on all
countries on a single model without any
warning [music] is highly questionable.
And the deemed export controls that also
restrict access to foreign nationals are
just absurd. He argued that we urgently
need a smart export control strategy
that denies adversaries access to
advanced technology while advantaging US
companies. But what's happening now is
the opposite. There's also this whole
discussion happening on Reddit and other
platforms about what this means for AI
development going forward. Users were
reporting that when they switched from
Fable 5 to Opus 4.8 mid-session, the new
model would sometimes still identify
itself as Fable 5 in commit messages,
which is this weird quirk where the
model gets confused by the previous
context. But the bigger concern people
are raising is what this means for
competition. A lot of folks are pointing
out that this basically hands China a
huge advantage in the AI race. If the US
is going to start pulling models offline
with no warning and restricting access
based on nationality, companies and
developers around the world are going to
start looking elsewhere. Chinese models
are getting cheaper and continuing to
catch up in capability. And users on
Reddit were saying they're canceling
their Claude subscriptions and switching
to alternatives [music] like Qwen 3.7.
Now, let's talk about the business side
of this because the timing is
particularly interesting.
Just last month, Anthropic announced
their annualized revenue had risen to
$47 billion
up from roughly $5 billion at their
September 2025 funding round. That's
more than doubling every quarter, which
economist Noah Smith called more
impressive revenue growth than any
company in history. [music] The
company's valuation has been absolutely
skyrocketing. In March 2025, they were
valued at $61.5 billion.
By September, that had tripled to 183
billion. By February 2026, it more than
doubled again to 380 billion. And by
May, investors were valuing the company
at $965 billion ahead of OpenAI at
roughly 852 billion. So, they're on the
verge of a potential trillion-dollar IPO
with attention already turning to them
as the next big candidate after SpaceX.
[music]
That's a 15.7 times increase in
valuation in little more than a year.
And while some people, including Eric
Reese who advises Anthropic on
governance, are reflexively skeptical
about these kinds of valuations, the
company's revenue growth actually makes
it look less crazy than it might seem at
first glance. Unlike SpaceX, which
trades at over 90 times revenues,
>> [music]
>> Anthropic at a trillion dollars would be
trading at around 20 times sales,
>> [music]
>> which isn't cheap, but it's relatively
conservative for a company still growing
at about 130% per [music] quarter. But,
here's the thing. All of that momentum
could be seriously disrupted by what's
happening with Fable 5. [music]
Even though the issue might get resolved
relatively quickly, this kind of
disruption is extremely distracting for
people inside the company.
Dan Shipper, CEO [music] of Every,
compared it to when Sam Altman got fired
from OpenAI, which was resolved
relatively quickly, but still disrupted
their momentum for a while. There's also
some really interesting data from
Anthropic's first public records survey
that adds context to all of this.
They surveyed nearly 52,000 Americans in
November and December 2025. And the
results show that 64% of Americans are
worried that AI will displace jobs.
>> [music]
>> That was the most common fear in every
single state. The second most prominent
fear was cognitive dependency at 56%
[music] where people worry that AI
integration will leave them unable to
think for themselves.
Job loss concerns are actually higher
among Americans with more education and
interestingly, people who use AI at work
every day are notably less worried about
job loss than people who don't use AI at
all. 54% versus 70%. When it comes to
trust, only 15% of Americans said they
trust AI companies to make decisions
about how the technology is developed
and used. [music] That was the lowest
figure for any institution they tested,
below even the federal government at
20%. [music]
But here's what's really important. 71%
of Americans say the government should
be involved in the development and
regulation of AI.
That's 79% among Democrats, 68% among
Republicans, and 69% among Independents.
It's a bipartisan supermajority. When
asked what should happen to ensure AI is
developed in humanity's interest,
Americans converged on two answers. Hold
AI companies legally liable for harm,
which 47% chose among their top three,
[music]
and prioritize safety over growth at
44%. So, the public clearly wants
government oversight and accountability.
But the question is whether this
specific action by the Trump
administration represents the kind of
transparent, fair, clear, and
technically grounded process [music]
that Anthropic and others are calling
for. Alan Rosenstein from the Lawfare
Institute suggested this could become
the first big First [music] Amendment AI
case, which would be a whole other
dimension to this story.
Anthropic says [music] it believes this
is a misunderstanding and that it is
working to restore access as soon as
possible. But honestly, the bigger issue
is not just whether Fable 5 comes back.
The bigger issue is the precedent this
sets. If a narrow potential jailbreak is
enough to pull a commercial AI model
offline, then almost every frontier
model is at risk. Every powerful model
can be jailbroken [music] in some
specific way. So, if the government can
say national security concern and remove
access based on limited evidence, then
we may have just seen the first real
limit on how much AI regular people are
allowed to use. And that is what makes
this serious and honestly kind of sad.
The model exists, the capability exists,
but once it becomes powerful enough, it
may no longer be something the public
can actually access.
Anthropic built strong safeguards,
worked with government agencies, and
spent thousands of hours testing Fable 5
before release. And even after all of
that, the model still got pulled. So,
this is no longer just about Fable 5. It
is about whether the public will
actually be allowed to use frontier AI
once it becomes powerful enough to
matter. So, what do you think?
Is this a necessary national security
move, or is this a dangerous precedent
for the future of public AI access?
Let me know your thoughts in the
comments below. And if you enjoyed the
video, make sure to like it, subscribe
for more AI news, and thanks for
watching. I'll catch you in the next
one.