Anthropic’s roughly 3,000 employees spent last Friday watching the company’s best jobs turn off, and nearly a week later, they still can’t explain why. The order from the White House was communicated less than 90 minutes later, and no one bothered to elaborate on the national security justification. Managers were told to alert customers about unexplained outages. Its two core models, Fable 5 and Mythos 5, were the most powerful Anthropic ever shipped, but they are now being used mid-task with no clear sense of what went wrong.So the workers did what people do when official channels go silent and took it to group chats. Their entries, later viewed by The New York Times, seem less like a crisis response and more like a room full of smart people trying to guess what happened to them. Explanations kept changing, the order made no sense, and silence from above only strengthened the theory. “Are we being bullied based on bad vibes?” one asked. Also, after a few days of silence, I fell into some dark place. “At what point do you feel like they don’t want us to exist?”
The explanation kept changing and that was the problem.
The whiplash started early. The reasons behind the closure will not remain silent. The first danger was that foreign companies could access the model. It was a vulnerability buried within them. Then, somehow, both. Engineers circulated contradictory news reports and asked in plain language what to believe.The order itself gave them no stability. This made the model inaccessible to all foreign nationals outside the United States, and even green card holders working within the United States. The net was so wide that the only way to follow it was to unplug everyone. Within 15 minutes of Friday’s call, executives were already on the phone with officials asking for the real reason. they got nothing.
Anthropic had to figure out the trigger themselves
Here are some details that show how tenuous communication was: No one in the government told Anthropic what caused this. The company figured out the problem on its own, came across a research paper, and then had to call the White House to ask if the paper was the cause.The paper comes from Amazon, and is one of the strange wrinkles in the whole affair, considering Amazon has pledged to invest up to $33 billion in Anthropic. The researchers showed that Fable 5 could be induced to report flaws in vulnerable code, and CEO Andy Jassy reportedly took the concerns directly to Treasury Secretary Scott Bessent. Officials who read the findings described the findings as “horrifying.” Cybersecurity experts who read the same content were unimpressed, pointing out that OpenAI’s latest model does exactly this and stays online. Anthropic says it had clear government approval to launch Fable in the first place.
Employee suspects this is personal rather than technical
The gap between what the government claims and what experts see has worsened the internal atmosphere. For many staff members, this doesn’t feel like a security review. It can be read as resentment.They have a shape to point to. This is the second time in six months that Anthropic has clashed with the Trump administration. The label had never been used against a U.S. company since Anthropic was branded a “supply chain risk” by Secretary of Defense Pete Hegseth in February in a defense contract dispute. And by Tuesday, more than 150 security experts had signed an open letter calling for the restrictions to be lifted, pointing out that Fable’s guardrails are so aggressive that they will become a punch line in the security world on launch day. Employees shared it with others as if to justify it.From this time on, the temperature has dropped a little. President Trump told Axios this week that he no longer views Anthropic as a threat and said after the G7 that CEO Dario Amodei was a “nice guy” and “smart.” But Monday and Tuesday’s meetings yielded no progress, with those actually building the models still stuck at the starting point, exchanging theories about why their work has taken a turn for the worse.
