Key Points
- Federal officials have rejected Anthropic’s proposal to grant Mythos AI access to 70 additional organizations beyond the current roster.
- Anthropic restricted Mythos from public availability because of its capacity to identify and leverage software security flaws.
- The current user base operates through “Glasswing,” a restricted program serving companies like Apple, Microsoft, and Nvidia.
- Federal authorities fear expanded distribution could diminish government computing resources needed for the AI system.
- Unapproved individuals obtained Mythos access on the same date Anthropic revealed its restricted distribution strategy.
In early April 2026, Anthropic introduced Mythos, presenting the AI model as a sophisticated system for identifying and exploiting security weaknesses in essential software platforms. According to the company, this technology demonstrates the ability to uncover security gaps that remained hidden for years, escaping detection by both human security professionals and existing automated scanning systems.
Given these powerful capabilities, Anthropic chose to withhold Mythos from general public distribution. The company established “Glasswing,” a controlled access initiative that provides a specialized version of the model to select organizations including Apple, Microsoft, and Nvidia for strengthening their security frameworks.
Anthropic subsequently submitted a proposal to extend access to approximately 70 more entities, creating a total user base of roughly 120 organizations. Federal officials have since moved to block this expansion.
Federal Opposition Centers on Security and Resources
According to Bloomberg News, a White House representative speaking without attribution confirmed that Trump administration officials have communicated their disapproval of the expansion proposal to Anthropic leadership.
Computing resource allocation stands as a primary concern. Federal authorities express apprehension that broadening Mythos distribution to additional companies might compromise government access to sufficient computing power for operating the technology, the Wall Street Journal reported.
A White House spokesperson informed the Journal that the administration seeks to harmonize technological advancement with security requirements while guaranteeing responsible AI deployment.
This opposition intensifies existing friction between the Trump administration and Anthropic. President Trump issued an executive directive in February prohibiting federal agencies from utilizing Anthropic’s products after Defense Secretary Pete Hegseth classified the organization as a national security supply chain threat.
Anthropic has filed legal action contesting these restrictions.
White House Meetings and Security Breach
The relationship between Anthropic and federal officials demonstrated potential progress during early April discussions. Dario Amodei, the company’s Chief Executive, participated in meetings with administration representatives at the White House, with Anthropic characterizing the exchanges as “productive.”
Complications emerged when Bloomberg News disclosed that unauthorized individuals within a private digital community obtained Mythos access coinciding with Anthropic’s initial announcement of its controlled distribution approach. Anthropic acknowledged an ongoing investigation into this breach.
Various observers have questioned whether Anthropic has inflated Mythos’s actual performance capabilities. The company faces intense competition, particularly from OpenAI, which pursues comparable technological development.
Anthropic representatives declined to provide statements regarding the federal opposition. White House officials did not respond to commentary requests.
