Woofun AI reports that Anthropic is restoring public access to its Claude Fable 5 and Mythos 5 models after the US government lifted export controls on Wednesday. These frontier AI systems had been offline since June 12, when authorities mandated an immediate suspension following reports that researchers successfully bypassed Fable 5's safety safeguards. The government intervention aimed to prevent potential national cybersecurity threats arising from malicious exploitation of the models' capabilities.

The redeployment follows a series of productive conversations between Anthropic and US officials, resulting in the implementation of new classifiers designed to block specific cybersecurity tasks. US Secretary of Commerce Howard Lutnick confirmed on X that the two-week collaboration ensured alignment across the US Government while strengthening America's leadership in AI. White House Chief of Staff Susie Wiles echoed this sentiment, stating the priority remains to deploy the best technology as quickly and safely as possible.

The initial restrictions were triggered by a report revealing that Amazon researchers identified a method to bypass Fable 5's safeguards, prompting the model to identify several software vulnerabilities. In response, Anthropic argued in a blog post that this risk was not unique to Fable 5, noting that weaker models could also identify the same vulnerabilities and produce identical exploits. A well-known AI researcher had previously claimed to jailbreak Fable 5 within 48 hours of its June launch, sharing screenshots demonstrating how he bypassed the model's safety guardrails before the government intervention.

Structurally, Anthropic is now drafting a consensus framework with Amazon, Microsoft, Google, and other partners under Project Glasswing to assess the severity of AI jailbreaks. This collaboration, announced in April, aims to safeguard against AI cybersecurity threats through standardized evaluation metrics. The company is simultaneously scaling up collaboration with the US government on AI model testing and safeguards, which will include pre-release access to models for evaluation and dedicated resources for joint research.

Experts and technologists had previously warned that the suspension raised concerns about state control over frontier AI technology and set a dangerous precedent. The resolution of this incident marks a critical juncture where regulatory oversight and rapid deployment must coexist without stifling innovation. This event underscores the ongoing tension between national security imperatives and the open development of advanced artificial intelligence.