US allows Anthropic partial release of Mythos AI to 100 approved organizations
The U.S. government has allowed artificial intelligence company Anthropic to release its most powerful Mythos 5 model to roughly 100 approved U.S. companies and federal agencies, ending a two-week standoff. In early June, the Trump administration blocked all access to Mythos 5 and its consumer version Fable 5 over national security concerns that the powerful cybersecurity tool could be misused to find computer vulnerabilities. After negotiations with Commerce Secretary Howard Lutnick, the government determined sufficient safeguards were in place to permit release to vetted organizations focused on cybersecurity and critical infrastructure defense. Anthropic will continue discussions with officials about restoring broader access to Fable 5. The decision reflects an emerging regulatory framework where the government maintains approval authority over frontier AI models, even as tech leaders debate whether such control slows American AI innovation relative to competitors like China.
Key facts
- Roughly 100 U.S. companies and federal agencies approved for Mythos 5 access as of June 26, 2026
- Original export ban imposed June 12 forced complete shutdown of both Mythos 5 and Fable 5 models
- Commerce Secretary Howard Lutnick cited 'significant progress' and 'appropriate safeguards' in approval letter
- Fable 5 consumer model remains unavailable; no timeline for its return announced
- Over 100 cybersecurity experts signed open letter challenging government's factual basis for restrictions
How different outlets framed it
Left-leaning coverage: Proponents of AI regulation view this as appropriate government oversight of powerful technology with dual-use risks. The decision balances innovation with national security, ensuring that advanced models are deployed responsibly to trusted defenders rather than left entirely to company discretion.
Right-leaning coverage: Critics on the right and among tech libertarians argue the government is overstepping by picking which companies can access technology, citing Stanford cybersecurity expert warnings that the restrictions lack factual basis and weaken U.S. competitive advantage against China. They contend AI developers need stability, not sudden shifting restrictions.
Limits of this briefing
The government has not publicly disclosed the specific safeguards that triggered approval, the selection criteria for approved organizations, or detailed technical concerns beyond general reference to potential jailbreaking risks. Media coverage relies heavily on official statements and lacks independent verification of the security claims underlying the initial restrictions.
Sources
- CNN Business: US government allows Anthropic limited release of AI model that sparked cybersecurity concerns Centre
- CNBC: Trump admin allows Anthropic to release Mythos AI model to some companies, government agencies Centre
- Semafor: Exclusive / US releases powerful Anthropic model Mythos to some US companies Centre
- TechPolicy.Press: Anthropic's Mythos Recall and the White House's Missing AI Safety Playbook Right
- ABC News: OpenAI limits its latest ChatGPT product to Trump-approved customers during cybersecurity review Centre