The Trump administration has partially rolled back its recent suspension of Anthropic’s most advanced artificial intelligence (AI) models, allowing it to grant access to its powerful Claude Mythos 5 system to more than 100 trusted U.S. organizations.
The decision, detailed in a letter from Commerce Secretary Howard Lutnick to Anthropic’s co-founder and chief compute officer Tom Brown, reverses a sweeping June 12 export control directive that had abruptly sidelined Anthropic’s top systems over national security anxieties.
Lutnick noted “significant progress” in developing appropriate safeguards justified the policy shift, permitting select Fortune 500 corporations, financial institutions, and government agencies to redeploy the software.
Anthropic confirmed the development late Friday, noting that Mythos 5 — its strongest cybersecurity model — will immediately return to service for entities operating and defending critical infrastructure. The restoration also lifts export licensing restrictions for non-U.S. citizen employees working within these vetted partner networks. However, Anthropic’s mass-consumer model, Fable 5, remains offline while discussions with regulators continue.
The administration’s aggressive oversight reflects deepening concerns that sophisticated frontier models could be weaponized by foreign adversaries like China and Russia to accelerate cyberattacks against critical economic sectors. Hours before Anthropic’s announcement, rival lab OpenAI revealed it was also delaying the full public launch of its new GPT-5.6 family at the government’s request, restricting initial access to a small, pre-vetted group of partners instead.
Despite the breakthrough, the administration’s secretive vetting process has drawn sharp criticism from industry leaders and civil liberties groups. Critics argue that the federal government is wielding excessive, non-transparent authority over commercial AI distribution.
OpenAI CEO Sam Altman expressed disapproval on social media, writing that while safety testing is vital, he disagreed with “the idea of the government picking the customers.”
Similarly, John Coleman, legislative counsel for the Foundation for Individual Rights and Expression, warned that the opaque selection process lacks transparency and undermines the rule of law.
The regulatory crackdown comes on the heels of a newly signed executive order by President Donald Trump, which establishes a voluntary framework giving the government up to 30 days to review “covered frontier models” before public release. The tense dynamic has been especially pronounced for Anthropic, which was recently placed on a national security blacklist after refusing to allow the U.S. military to use its technology for domestic surveillance and fully autonomous weaponry.
Industry analysts warn that prolonged regulatory gridlock could ultimately harm domestic interests.
Kate Koren, an analyst at the Center for Strategic and International Studies and a former Commerce Department official, called the selective rollback a practical interim step but warned of broader consequences. “The longer there isn’t a system in place that will allow U.S. companies to widely release new models, the more likely it is that China will be able to catch up, Koren said.
