WASHINGTON, June 27 (Parliament Politics Magazine) – The U.S. government has partially reversed an earlier suspension order, allowing Anthropic to provide its Claude Mythos 5 artificial intelligence model to a specific list of approved organizations. This decision comes just two weeks after federal officials ordered a complete shutdown of the system due to national security concerns regarding potential misuse by foreign intelligence actors.
More than 100 U.S. companies and institutions are now permitted to regain access to the powerful cybersecurity tool. The list primarily includes federal agencies and Fortune 500 corporations that manage critical infrastructure. Commerce Secretary Howard Lutnick confirmed that sufficient security safeguards are now in place to allow the deployment of Mythos AI, removing the previous requirement for specific export licenses for these entities and their foreign national employees.
Many of the companies receiving access are participants in Project Glasswing, an initiative involving various technology leaders. While Anthropic moves to restore services for these approved partners, the consumer-facing version, Fable 5, remains under restricted status as negotiations continue.
The federal government previously blocked these advanced models over fears that they could be jailbroken to accelerate sophisticated cyberattacks against sensitive sectors like banking. The recent shift in policy highlights an evolving landscape where frontier models are increasingly treated as strategic national assets rather than standard commercial software.
This regulatory tightening is part of a broader push by the administration to oversee AI deployment. On the same day Anthropic received its partial reversal, rival firm OpenAI announced it would delay the public release of its GPT-5.6 model at the request of federal authorities, limiting access to a similarly vetted group of partners.
The government’s decision to curate a list of authorized users has drawn significant criticism regarding transparency and the potential for federal overreach. Skeptics argue that the selection process lacks clear criteria, raising questions about how such power is wielded over private innovation.
“No one knows how these companies are picked and why everyone else is excluded,” said John Coleman, legislative counsel for the Foundation for Individual Rights and Expression.
The friction between the government and AI labs is further complicated by past tensions. Anthropic previously drew scrutiny after reportedly resisting requests to incorporate its models into military surveillance and autonomous weapons programs. While the current directive serves as a functional interim measure, analysts warn that the lack of a standardized process for public release could have long-term consequences for the domestic industry’s competitive edge against foreign counterparts.
Both Anthropic and OpenAI are navigating these requirements while preparing for future public offerings, forced to balance federal safety mandates with the need for broader product availability. For now, the focus remains on ensuring that the most capable cybersecurity tools remain within the reach of defensive infrastructure while limiting the risks of systemic exploitation.
