Trump Administration Orders Anthropic to Disable Advanced AI Models Over SK Telecom's Alleged China Ties

The Trump administration ordered AI firm Anthropic to disable its advanced Fable 5 and Mythos 5 models, citing national security risks. The directive followed concerns that a South Korean partner, SK Telecom, had alleged China ties and that a "jailbreak" vulnerability, reportedly flagged by investor Amazon, could expose the models' cyber capabilities. Anthropic complied, sparking industry debate about government overreach.

Anthropic model shutdown order

• ▪The Trump administration ordered Anthropic to suspend access to its Fable 5 and Mythos 5 AI models for all foreign nationals, citing national security concerns
• ▪Lutnick's letter warned Anthropic of potential criminal and civil penalties if the company failed to comply with the export control order
• ▪Anthropic's other models, including Opus, Sonnet, and Haiku, were not affected by the order and remain available
• ▪To ensure compliance, Anthropic disabled Fable 5 and Mythos 5 for all customers worldwide
• ▪The directive was issued on June 12, 2026, by Commerce Secretary Howard Lutnick, who gave Anthropic less than 90 minutes to comply

Fable 5 jailbreak vulnerability

• ▪The order was partly triggered by a reported "jailbreak" that could bypass Fable 5's safety guardrails, potentially turning it into an unrestricted cyber tool
• ▪The company argued that applying a "perfect jailbreak resistance" standard would halt all new model deployments across the industry
• ▪Anthropic disputed the severity of the jailbreak, stating the capability demonstrated is available from other public models like OpenAI’s GPT-5.5
• ▪White House AI adviser David Sacks claimed a trusted partner found the jailbreak and that Anthropic CEO Dario Amodei refused to fix it or de-deploy the model

Amazon's role triggering crackdown

• ▪Amazon is a major investor in Anthropic, having committed billions of dollars, and provides much of its cloud infrastructure
• ▪Multiple reports identified Amazon as the "trusted partner" that flagged the Fable 5 jailbreak to the U.S. government
• ▪Amazon CEO Andy Jassy reportedly told officials that his company's researchers used Fable 5 to obtain information that could be used in cyberattacks

Anthropic-Trump administration conflict

• ▪The U.S. government's action was also triggered by concerns over Anthropic granting Mythos 5 access to SK Telecom, a South Korean firm alleged to have ties to China
• ▪SK Telecom, which invested $100 million in Anthropic in 2023, denied having ties to China
• ▪Days before the shutdown, Anthropic CEO Dario Amodei published an essay calling for government authority to block AI models that fail safety tests
• ▪Before the blanket export control order, the White House asked Anthropic to revoke SK Telecom's access to Mythos, and the company complied
• ▪The Department of Defense had previously designated Anthropic a "supply chain risk" amid disputes over contract terms for military use of its AI

Export control precedent

• ▪The shutdown occurred as Anthropic was preparing for an IPO, after its annual recurring revenue grew from $9 billion to $47 billion in about six months
• ▪The directive is considered one of the most aggressive uses of export-control powers against a commercially deployed AI model
• ▪An Anthropic executive stated the "jailbreak" is common in recent AI models and expected the company's models to be usable again within days