Leaked internal chats from Anthropic show staff wrestling with sudden export controls that forced a global shutdown of their freshly launched Fable 5 and Mythos 5 models, just days after rollout. Employees voiced confusion over shifting explanations from leadership while a roughly $200 million Defense Department contract hung in the balance, amplifying worries that the company was being singled out.
the nytimes really didn't hold back on getting internal chat messages from anthropic
where the "same people" who previously claimed the ability to bring about a cybersecurity "reckoning"
are now left wondering if they're being "picked on, bullied, unfairly targeted" by the administration
and are also leaking these chats to the press
-----
nothing like some tough times to see who abandons ship first
Staff messages flagged potential damage to the planned 2026 IPO from the contract friction and model restrictions, though no revised timeline or valuation impact has surfaced yet.
Fable 5’s protections were mocked by security researchers for being overly strict before the government stepped in, leaving open how quickly access might return even if policy talks advance.
Negative users accuse Anthropic of hypocrisy for pushing AI safety fears and regulation then complaining about resulting government scrutiny, while a few defend the company as genuinely safety-focused.
No Digg Deeper questions have been answered for this story yet.
Very interesting: The NYT obtained and reported on internal Anthropic messages.
The message: Employees feel "unfairly targeted" and "feel like they (the US government) don't want us to exist."
However, the report omits the fact that Anthropic itself has repeatedly emphasized the security risks its model poses.
Messages here:
the nytimes really didn't hold back on getting internal chat messages from anthropic
where the "same people" who previously claimed the ability to bring about a cybersecurity "reckoning"
are now left wondering if they're being "picked on, bullied, unfairly targeted" by the administration
and are also leaking these chats to the press
-----
nothing like some tough times to see who abandons ship first
everyone at anthropic passed the culture fit interview of being ok if their stock went to 0 so clearly the leaks are sanctioned and they're just trying to appear relatable to win some public sympathy points
nyt article: https://www.nytimes.com/2026/06/17/technology/anthropic-trump-administration-fable.html
nyt article: https://www.nytimes.com/2026/06/17/technology/anthropic-trump-administration-fable.html
the nytimes really didn't hold back on getting internal chat messages from anthropic
where the "same people" who previously claimed the ability to bring about a cybersecurity "reckoning"
are now left wondering if they're being "picked on, bullied, unfairly targeted" by the administration
and are also leaking these chats to the press
-----
nothing like some tough times to see who abandons ship first
@suchenzang humour, or deep resentment?
the nytimes really didn't hold back on getting internal chat messages from anthropic
where the "same people" who previously claimed the ability to bring about a cybersecurity "reckoning"
are now left wondering if they're being "picked on, bullied, unfairly targeted" by the administration
and are also leaking these chats to the press
-----
nothing like some tough times to see who abandons ship first
@suchenzang @yacineMTB This was also my thought exactly; they are very clear about this position.
everyone at anthropic passed the culture fit interview of being ok if their stock went to 0 so clearly the leaks are sanctioned and they're just trying to appear relatable to win some public sympathy points
@kimmonismus “Safety” theater needs to end.
The government is holding AI back so they can choose who gets access.
Period.
This is 100% about control and money.
None of this is about safety. Real safety is possible.
Go use FreeLattice. The code is the proof. The platform is the proof. •
@suchenzang so i accidentally called it, huh?
1) having a victim mindset isn't going to help get anyone out of this pickle, especially for a place whose original founding principle was supposedly to avoid excessive commercialization
2) the fact any of this is being leaked to the press in real time is also unusual for a place that prizes its "culture fit" above all else
@Chaos2Cured open source is the key
@suchenzang Everyone is morally aligned until the stock actually goes to zero
@suchenzang @mattyryze i agree that people shouldn’t be leaking this stuff to the press, but very strange to call staff’s private questioning of the admin’s (obviously somewhat dubious) motivations a “victim mindset”
@Web3_0_AMATW HOLY MOLY
@kimmonismus
@kimmonismus The issues is: Antrophic just used fear mongering to advertise their model, and the low IQ Trump admin took their word for it.
@suchenzang not sure I get it, the employee quotes in your screenshots seem extremely reasonable given the circumstances, and don't undermine Anthropic in any way that I can see
in what way does this "nothing like some tough times to see who abandons ship first" apply
@suchenzang I mean compare to how many leaks there were in November 2023 when Twitter at large found out in real time the extent to which Sam Altman is a conniving compulsive liar, OpenAI was a sieve
@mattyryze @suchenzang The anti-Anthropic mindset is that they are evil masterminds withholding the world's most destructive power but also they are exaggerating the destructive power of AI to try to force regulation on their competitors.
@suchenzang "if nyt is reputable and validates all of its sources, it is definitely not reporting off of a single leak" having read the article just now that is not obvious to me, they had an internal chat log(s) and they would of course validate it's authenticity
@suchenzang @yacineMTB Yep! The (somewhat, depends) nice part is that from their recent actions, it does seem like it's not just a motto but at least leadership seems to really mean it.
Holy shit, we haven't built the real safety, that was my initial point. Cybersecurity firms need to be given advantage access to frontier models and shit tons of reserved gpus. A Cybersecurity security apocalypse isn't a llm going down for a week, its the *internet* going down for who knows how long, and then the normies would want your blood, they would literally want to put you in prison for espousing pro open source ideas.
