Newsletters
On Monday, Anthropic officials and representatives of the Trump administration met in Washington.
The U.S. AI company is trying to overturn restrictions on two of its most advanced models, according to several reports.
The restrictions, introduced on Friday, prevent foreign governments, companies, and individuals, including foreign nationals working at Anthropic in the U.S., from accessing Anthropic's Fable 5 and Mythos 5 models for national security reasons.
Anthropic said it received a government order on Friday to suspend access to both models, just three days after Fable 5 was launched to the public.
The company believes the order was due to concerns that Fable 5 could be 'jailbroken,' meaning it could be manipulated to bypass safety measures meant to prevent harmful uses.
A person familiar with the talks said the two sides have not yet agreed to lift the restrictions after Monday's meeting. An Anthropic spokesperson said both groups are working quickly to get this resolved.'
'This is part of our ongoing commitment to work with the administration toward our shared goal of protecting U.S. critical infrastructure and maintaining the U.S. lead in cyber defense,' the spokesperson added.
According to The Wall Street Journal, Anthropic officials went to Washington over the weekend before the meeting. The talks included Commerce Secretary Howard Lutnick, National Cyber Director Sean Cairncross, Anthropic co-founder Tom Brown, and Sarah Heck, who leads public policy at the company.
2 min read
Anthropic first announced Mythos 5 in April but decided not to release it to the public, citing concerns about its ability to find software vulnerabilities, which are weaknesses in code that could be exploited by bad actors.
Last Tuesday, the company launched Mythos 5 for select users and released Fable 5, a version with additional safety features intended for broader public use.
The restrictions were put in place just days after the launch, a sudden change for a model that Anthropic had promoted as safer and publicly available.
The government's decision was criticized by cybersecurity executives, researchers, and technology experts, who signed an open letter asking for the restrictions to be removed.
They argued that Fable 5 and Mythos 5 do not have unique features that other advanced AI systems lack.
They also warned that limiting access could hurt US leadership in AI and weaken the country's cyber defense, and they called for future AI risk assessments to be more transparent and scientific.
The restrictions come against the backdrop of an ongoing dispute between Anthropic and the US government over the use of advanced AI for military purposes, citing concerns about mass surveillance of US citizens and autonomous weapons systems.
That position led the Pentagon to cut contracts with the company earlier this year and prompted President Donald Trump to direct federal agencies to halt use of Anthropic technology.
The Pentagon subsequently designated the company a procurement-security risk.
Anthropic later sued the US government, alleging that officials had illegally retaliated against the company for its position on AI safety.
According to reports, Amazon CEO Andy Jassy was one of the tech leaders who privately warned administration officials about security risks in Anthropic's most advanced models.
White House AI adviser David Sacks said the new restrictions are not related to the earlier Pentagon dispute. Defense Secretary Pete Hegseth wrote on X that the Pentagon's earlier decision to end its relationship with Anthropic was the right choice.
This month, Trump signed an executive order establishing a voluntary federal review of national security risks in advanced AI models before they are released. This measure was introduced due to broader concerns about what Mythos 5 can do.
The current dispute has intensified debate in Washington over how to regulate advanced AI models without undermining US technological competitiveness.
Set as preferred source
Set as preferred source
1 min read
