US Eases Restrictions, Allows ‘Trusted’ Organisations Access to Anthropic’s Powerful Mythos AI Model

Free Speech Advocates, Tech Experts Say AI Access Selection Process Lacks Transparency

Web Desk

WASHINGTON: The United States has partially lifted restrictions on access to Anthropic’s advanced artificial intelligence model, Mythos 5, allowing a limited number of “trusted” American organisations to use the technology.

According to Anthropic, more than 100 companies and institutions, including several Fortune 500 firms, will regain access to the AI system after the US government revised an earlier decision that had suspended its deployment over national security concerns.

The move comes amid growing concerns in Washington that highly advanced AI tools could be misused by hostile actors or foreign intelligence agencies. As a result, the administration has adopted a cautious approach toward the release of next-generation AI models developed by leading companies such as Anthropic and OpenAI.

Anthropic said the government has authorised the redeployment of Mythos 5 to organisations responsible for operating and protecting critical infrastructure in the United States.

"We are rapidly restoring access for these organisations and will continue working with the government to broaden availability of Mythos 5 and eventually return Fable 5 to general use," the company said in a statement.

The government's role in deciding which companies receive access has sparked criticism from free speech advocates and technology experts, who argue that the selection process lacks transparency.

Meanwhile, OpenAI recently announced that it would delay the wider release of GPT-5.6 at the request of US authorities, limiting access to a small group of carefully vetted partners.

Security experts have warned that powerful AI systems, if misused, could significantly enhance sophisticated cyberattacks, particularly against sectors that rely on complex and aging digital infrastructure.

US officials said progress had been made in addressing potential risks associated with these advanced AI models, although specific safeguards introduced by the companies have not been publicly disclosed.