Claude users hoping to continue testing out the app’s latest artificial intelligence models, Fable and Mythos, are starting their weekends upset after the U.S. government ordered the company to suspend access.
On Friday, the U.S. government issued an export control directive suspending all access to Anthropic’s Fable 5 and Mythos 5 models by any foreign national inside or outside the U.S. This included Anthropic’s own employees. The company said that in order to comply with the order, it would have to suspend access to the models to all customers.
The order couldn’t have come at a worse time for the company, as it just filed paperwork for a proposed initial public offering of its common stock. Investors are watching as other major AI companies, like SpaceX and OpenAI, have also announced their IPO ambitions.
Why did the government block Mythos?
Anthropic said the government’s order did not specify the national security concern. But the company said it believes the Trump administration became aware of a way people can bypass security protocols on the Fable 5 model. Anthropic described the vulnerabilities as “previously known” and “minor.”
“These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass,” the company wrote in a blog post.
The company said it was deliberate with how it released Fable and Mythos, even providing major corporations and the government with early access to the model to test its usage.
Anthropic said its team instituted “strong safeguards” that reduce the likelihood that people can misuse Fable for tasks related to cybersecurity. It said its safeguards were so robust that “many users have complained that they are overly broad.”
Company officials said no one has been able to find a universal method to “jailbreak,” or break through those safeguards, which would unblock a large array of cyber capabilities.
According to Axios, Commerce Secretary Howard Lutnick sent a letter to Anthropic CEO Dario Amodei on Friday, telling him that the models would be subject to export controls to any location outside the U.S. and to all foreigners within the country.
The outlet said that the administration had tried to get Anthropic to pause the models’ release but was unsuccessful. An official told Axios that the model would remain locked until the U.S. government’s national security apparatus is stronger. They said that it could take a few weeks.
Bad timing for Anthropic
Less than two weeks before this order came out, Anthropic filed to go public after closing a $65 billion financing round that valued the company at nearly $1 trillion.
The new order could hurt that deal as the regulatory uncertainty could become a product risk for investors. The export control decision could make investors less enthusiastic about the prospect and cause them to question whether Anthropic can stay at the bleeding-edge of AI development.
Politics may also be playing a role in the decision, since several members of the Trump administration have said Anthropic is “woke” and “leftist.”
Anthropic’s history with the Trump admin
It’s ironic for Anthropic, which prides itself on being the more ethical AI company, to be the only AI company subject to government regulations on what it releases.
The company’s ethics even got it into a fight with the U.S. military after it refused to give the Pentagon full control over its AI models, citing its own rules on AI weapons use. This caused the military to blacklist Anthropic and prohibit any companies connected with the U.S. government from accessing it.
While complying with the order and previously agreeing that the government should be able to block unsafe deployments, they disagree with the Trump administration. Officials said that if the government used the same reasoning for all major AI companies, “it would essentially halt all new model deployments for all frontier model providers.”
“As we have stated publicly, we believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts,” Anthropic said. “This action does not adhere to those principles.”
Round out your reading
- People are overdosing on GLP-1 injections. What does that look like?
- Researchers put chatbots in a simulation. Grok ended the world in 4 days.
- Fearing more e-bike injuries, Illinois looks to become 10th state to mandate registration, insurance.
- Why are young professionals facing such a tough job market? Hint: It’s not just AI.
- We’re building a new Straight Arrow. Help us shape our future by taking our survey.