Defense Secretary Pete Hegseth took a shot at Anthropic on Saturday, after the artificial intelligence company discontinued access to two of its models to comply with a directive from the Trump administration.<\/p>\n\n\n\n
\u201cThree months ago, @DeptofWar kicked @AnthropicAI out of our building\u2014forever. Every passing day proves why that was the right move,\u201d Hegseth wrote on the social platform X.<\/p>\n\n\n\n
Earlier this year, Hegseth labeled Anthropic as a supply chain risk and prohibited the use of its Claude AI chat bot within the Pentagon. Anthropic has sued the administration over the designation, which came after the firm\u2019s CEO, Dario Amodei, sought to ensure that the government would not use the tool for fully autonomous lethal weapons or mass domestic surveillance of Americans.<\/p>\n\n\n\n
The latest clash between Anthropic and the administration was over the company\u2019s Fable 5 and Mythos 5 models. After the administration issued an export control directive to \u201csuspend all access\u201d to the models by foreign nationals, Anthropic said Friday that it had to disable access to Fable 5 and Mythos 5 for all customers to comply.<\/p>\n\n\n\n
In a statement, the company noted that it believes that the government issued the order after discovering a method of \u201cbypassing\u201d Fable 5, also known as \u201cjailbreaking.\u201d<\/p>\n\n\n\n
Anthropic added, \u201cWe reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities. 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.\u201d<\/p>\n\n\n\n
A source close to Anthropic \u2014 which is looking to go public later this year with a potential $1 trillion valuation \u2014 told The Hill that the administration informed the company that it had 90 minutes to pull Fable 5 on Friday, without providing any prior indication regarding a threat to national security.<\/p>\n\n\n\n
The source noted that Amodei and senior company personnel were in touch with the White House within 15 minutes of the initial order, and the administration did not provide the firm with any details regarding the directive.<\/p>\n\n\n\n
Earlier this month, President Trump signed an executive order stating that AI labs can voluntarily provide the government with their models for a period of up to 30 days before they plan to release them publicly.<\/p>\n\n\n\n
Amodei backed the order earlier this week and proposed mandatory testing of AI models by a third-party auditor to check for risks related to cybersecurity, biological weapons, loss of control of AI systems and automated research and development.<\/p>\n\n\n\n
But on Friday, Anthropic said that the latest directive from the administration was unnecessary and would \u201cessentially halt all new model deployments for all frontier model providers\u201d across the AI industry.<\/p>\n\n\n\n
\u201cAs 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,\u201d the company stated. \u201cThis action does not adhere to those principles.\u201d<\/p>\n\n\n\n
The source close to the company noted that it worked with relevant government agencies on pre-release testing of Fable 5 and thereafter received an explicit approval approval to deploy the model. The source also stated that Anthropic is eager to resolve the issue with the administration.<\/p>\n\n\n\n
But David Sacks, a special adviser to Trump on AI and cryptocurrency from January 2025 to this past March, said Saturday that Anthropic\u2019s response to the directive \u201cis very much at odds with their branding and ethos as a safe AI research\u201d firm.<\/p>\n\n\n\n
\u201cThe Admin\u2019s hope now is that Anthropic remediates the safety issue, the export control is lifted, and Fable goes back into general release,\u201d Sacks wrote on X. \u201cThe Admin wants all of this to happen as soon as possible. It is frankly bewildered that Anthropic hasn\u2019t wanted to comply with safety requests that it previously said were its highest priority.\u201d<\/p>\n\n\n\n
Administration officials and allies are not alone in their support of the order, however. Democratic Sen. Mark Kelly (Ariz.) said Sunday that issuing the directive was the right call, and AI companies should be \u201cincredibly careful\u201d when releasing their models.<\/p>\n\n\n\n
\u201cI think Anthropic is a good example, [it] seems to be willing to work with the federal government on this to make sure that we do not make a mistake and release something that we will later regret,\u201d he told host Margaret Brennan on CBS News\u2019s \u201cFace the Nation.\u201d<\/p>\n\n\n\n