In short
- The US government ordered Anthropic to suspend access to two of its most advanced AI models, Fable 5 and Mythos 5, citing national security concerns due to the risk of incarceration.
- Anthropic followed suit but disputed its findings, arguing that the vulnerability was simple and could be exploited using other publicly available versions such as GPT-5.5.
- The company warned that the guidelines would set a dangerous precedent that, if applied across the board, could halt the deployment of new AI models.
The US government issued an emergency export ban on Friday ordering Anthropic stopping at the same time to get his two most powerful types of artificial intelligence—Claude Fable 5 and Claude Mythos 5, both. it was released a few days earlier– for all foreign nationals, including company employees, citing national security.
This law prohibits a foreign national from acquiring models, either inside or outside the United States. The scale of the system forced Anthropic to ban all of its customers’ brands to ensure compliance.
The letter did not go into detail about national security, but the government believes it has identified a way to bypass, or “jailbreak,” the publicly available version of Fable 5. Mythos 5, which has fewer barriers and is more powerful in detecting cybersecurity issues, was found in the selection of partners.
Anthropic disputed the seriousness of the findings. The company said it reviewed the technology demo and confirmed that the identified vulnerabilities appear to be simple, and that other publicly available models can also be found without requiring a bypass.
The company said that so far the government has only provided verbal evidence of the narrow, non-universal jailbreak – mostly asking that the prototype read a specific codebase and fix any software bugs. Anthropic added that it confirmed that a large number of data points are already available from competing models, including OpenAI’s GPT-5.5.
While following the guidelines, Anthropic said it believes the incident sets a dangerous precedent. The company wrote: “If this standard were to be applied across the board, we believe it would stop the shipment of all models to all sample providers.”
Access to all other Anthropic species will not be affected. The company said it is working to restore access as soon as possible.
I have had several conversations with people inside and outside the government about the current situation with Anthropic, and here is what I believe to be true:
– As we know, Anthropic released its Mythos series earlier this week under the trade name Fable.…
— David Sacks (@DavidSacks) June 13, 2026
On Saturday, David Sacks-chairman of the Council of Advisers on Science and Technology-wrote on X that “a trusted and trusted friend of Anthropic and the US government who was testing Fable came up with the prison of the guards. The manager asked (Anthropic CEO Dario Amodei) to fix the problem of the prison or remove the sample. Dario refused.”
Sacks also said that Anthropic’s actions on the government’s request contradicted the company’s public statements on the need for AI security and control — which Amodei also did. shared this week in a blog post.
“Anthropic prioritized consumer brand continuity over security,” Sacks wrote. “In doing so, the administrator granted the export authority. The administrator did this reluctantly. It is very surprising that Anthropic did not want to comply with a reasonable security request (for example, to fix the prison problem). What Anthropic does is completely against their brand and their culture as a safe AI research group.”
Sacks added that management hopes that Anthropic will fix the issue and that Fable 5 may return to public release.
“Management appreciates Anthropic’s expertise and feels that this issue, while serious, should be easily resolved,” he wrote. “The ball is in Anthropic’s court.”
Earlier this year, Anthropic and the US government made headlines when the AI company refused to sign an additional agreement to allow American surveillance and autonomous weapons.
President Donald Trump he sued the company Negotiations ended, and the Department of Defense labeled Anthropic a “disposable resource” — a designation the company has. was challenged in court. Since then, there have been reports of tensions between the two sides he has faded when the government wanted to use Claude Mythos and other examples.
Daily Debrief A letter
Start each day with top stories right here, including originals, podcasts, videos and more.
