Anthropic's most advanced AI models could be restored shortly following dramatic emergency freeze

At a glance:

• US emergency directive forced Anthropic to suspend Fable 5 and Mythos 5 globally.
• Anthropic’s international director Chris Ciauri says the models could return within days, expressing confidence.
• The freeze arose from concerns that South Korean telecom SK Telecom was added to Project Glasswing over alleged China ties.

Emergency directive and model freeze

Anthropic was compelled to pull its flagship Fable 5 and Mythos 5 AI models worldwide after the U.S. government issued an emergency national security directive last week. The directive targets advanced systems that can scan, identify, and patch critical code vulnerabilities, raising fears that they could be weaponized against infrastructure. While paying customers remain locked out, the company has not disclosed an exact restoration date, only hinting at a near‑future release.

The press conference in Seoul, led by Managing Director Chris Ciauri, emphasized Anthropic’s confidence that the models will be back online shortly. Ciauri stated that the company is “very confident that in the coming days, the models will become available again,” as reported by Korea JoongAng Daily. This optimism contrasts with the sudden shutdown that disrupted global access to the models.

Project Glasswing and partners

Project Glasswing is an exclusive corporate cybersecurity initiative that grants select entities access to Mythos 5, a model designed to audit and repair deeply hidden vulnerabilities in complex codebases. Anthropic says the program currently involves approximately 150 partners, including Google, NVIDIA, Microsoft, and Apple. These partners receive early, controlled access to test the model’s defensive capabilities while mitigating misuse risks.

• Google
• NVIDIA
• Microsoft
• Apple

The partnership list illustrates the high‑profile nature of the initiative, bringing together major tech firms that collaborate on threat‑intelligence sharing. However, the U.S. government’s export controls emerged after a dispute over SK Telecom’s inclusion in the program, citing alleged ties to China, a claim the Korean carrier denies. The dispute highlighted the geopolitical sensitivities surrounding AI development.

Outlook and market impact

Paying customers now await clarification on whether Anthropic’s upcoming regional guardrails and security patches will satisfy Washington’s concerns. The company must balance rapid model reinstatement with compliance to new export‑control requirements, which could set a precedent for other AI firms. Analysts warn that the episode may accelerate regulatory scrutiny of powerful AI capabilities across the industry.

The episode also highlights the broader tension between national security and AI innovation, as governments grapple with how to manage technologies that can both protect and threaten critical infrastructure. Observers will watch how quickly Anthropic can implement the required safeguards and whether the restored models will retain full functionality under the new oversight framework. Such decisions could influence future export policies for AI research worldwide.