Anthropic disabled two newly released artificial intelligence models on Friday night following a directive from the US Commerce Department, marking an unusual intervention in the commercial AI sector. The company removed access to Fable 5 and Mythos 5 for all users after receiving export control restrictions that limit the models' availability to United States-based customers only.
The abrupt shutdown occurred just days after the models went live. In a statement released Friday evening, Anthropic explained that complying with the government order "requires that we must immediately disable Fable 5 and Mythos 5 for all our customers." Other Anthropic models remain operational and unaffected by the directive.
Security Concerns Trigger Government Action
According to Ars Technica AI, administration officials cited concerns about a vulnerability that circumvents safety mechanisms built into Fable 5. The reported jailbreak allows users to bypass classifier-based safeguards designed to prevent the model from responding to requests related to cybersecurity, chemistry, and biology topics. These fields are traditionally considered sensitive from a national security perspective.
The Trump administration requested a temporary pause on deploying these systems to allow government agencies time to strengthen defenses against similar vulnerabilities. An unnamed official told Axios that remediation efforts could be completed within weeks, suggesting the suspension may not be permanent.
What This Means for AI Governance
The incident represents a notable escalation in government oversight of large language model releases. Rather than pursuing regulatory frameworks through legislative channels, the administration wielded existing export control authorities to rapidly halt deployment of commercial AI systems deemed problematic.
- The shutdown affects only new models with reported security gaps
- Export controls restrict international access while domestic versions remain unavailable
- The government timeline suggests a targeted intervention rather than broad industry restrictions
Anthropic's quick compliance with the directive demonstrates how even leading AI companies must navigate evolving government authority over model deployment. The company did not publicly dispute the order or request extensions, indicating either agreement with safety concerns or recognition of the government's legal standing to enforce restrictions.
Implications for Competitive Landscape
The move creates uncertainty for competitors developing similar models. Other AI labs may face comparable scrutiny if their systems contain exploitable safety vulnerabilities. The incident also highlights tensions between rapid innovation cycles and government capacity to assess emerging risks.
The timing proves significant given ongoing debates about AI regulation. While some policymakers advocate for legislative guardrails, this episode shows federal agencies already possess enforcement tools through existing export control frameworks. The Commerce Department's willingness to deploy these authorities could set precedent for future interventions.
Industry observers will monitor whether government concerns about Fable 5 and Mythos 5 lead to broader restrictions on large language model development or whether the intervention remains narrowly tailored to specific vulnerability classes. The coming weeks will determine whether the administration's estimated timeline for remediation proves accurate.