Trump Administration Orders Total Block on Advanced US AI Technology

Anthropic has once again caused a stir due to new friction with the Donald Trump administration. The supplier of advanced AI tools for the US military has clashed with the government several times previously. Trump was angered when Anthropic refused to allow its AI tools to be used for creating autonomous weapons or spying on citizens, leading to a ban on the use of Anthropic’s AI tools across all federal agencies. The Department of Defense even placed Anthropic on a supply chain risk list, though tensions eased after Anthropic met directly with Trump at the White House. Shortly thereafter, Anthropic caused another uproar by launching the super-advanced AI model ‘Mythos’. This model can detect system vulnerabilities unknown to administrators, raising fears it could be exploited by cybercriminals to launch attack campaigns. Countries worldwide have warned the banking industry to strengthen data defence systems following Mythos’s launch. Upon its official release, the Trump administration immediately ordered that Mythos be banned for use by foreign entities. Anthropic responded by fully blocking Mythos services for all users. The company stated it received direct instructions regarding the export restriction on Friday, 12 June, and said the US government requested that the latest Claude models, Fable 5 and Mythos 5, be suspended for users of foreign nationality, both inside and outside US domicile, including Anthropic employees who are foreign nationals. The Trump administration claimed the order was to prevent national security risks but provided no detailed reasoning, according to Anthropic. The directive was issued just days after Anthropic CEO Dario Amodei published an essay advocating for ‘more serious and binding AI regulation’, including the ability to block models deemed unsafe. Amodei wrote that cutting-edge AI models, like aircraft, should undergo testing and technical audits, and their release should be blocked or cancelled as a threat to public safety if they fail to meet high safety standards. This is the latest public statement from Anthropic pushing for greater government oversight of the rapidly growing AI industry. Since splitting from OpenAI to start Anthropic in 2021, Amodei and his top executives have been staunch supporters of AI regulation and have backed legislation at state and federal levels. The company praised an AI executive order signed by President Trump earlier this month as an ‘important step’. However, the action on Friday was not the type of oversight Anthropic desired. The company said in its response that it disagreed that the administration’s findings warranted a product withdrawal, describing the disruption as a ‘misunderstanding’. Anthropic stated it believes the government should have the ability to block unsafe deployment as part of a transparent, fair, clear, and technically factual legal process, adding that this action did not align with those principles. Senior Anthropic employees flew to Washington, D.C., to meet with members of the Trump administration on Monday, 15 June. The company said both parties are working quickly to resolve the issue but has not indicated when its models will be operational again. Trump’s executive order, issued 10 days before the directive to Anthropic, lacked specific details. It asked companies to voluntarily provide models to the federal government to assess their capabilities before full release and gave administration officials 60 days to develop a relevant review framework. There is no indication the forced suspension of Anthropic’s latest models is related to that executive order. Instead, according to The Wall Street Journal, the move was triggered by conversations Amazon CEO Andy Jassy had with US officials, including Treasury Secretary Scott Bessent. Citing people familiar with the matter, the WSJ reported Jassy informed the administration that Amazon researchers used a series of prompts to make Fable 5 present information that could aid cyberattacks. Daniel Remler, a senior researcher at the Center for a New American Security, noted that while the president’s executive order called for voluntary compliance, the Friday directive had a very different tone, appearing as a mandatory order with potential consequences for non-compliance. Fable 5 and Mythos 5 are built upon the Claude Mythos Preview release from April, a powerful offering excelling at identifying security vulnerabilities in software. Anthropic limited its launch to a select group of companies as part of a cybersecurity initiative called Project Glasswing. Mythos 5 remains restricted to certain user groups, but Anthropic made Fable 5 available to its enterprise and paid customers.