U.S. Lifts Export Restrictions on Anthropic’s Claude Fable 5

The Department of Commerce intervention began on June 12, when officials ordered Anthropic to restrict access to the models for all foreign nationals. The directive stemmed from concerns that a potential jailbreak could enable the AI to assist in identifying software vulnerabilities. Anthropic initially resisted the recall, arguing that the reported security flaw was narrow and common among existing language models. The company warned that using such isolated incidents as a justification for withdrawing commercial technology could effectively halt all frontier model deployments across the industry.

To secure the green light for redeployment, Anthropic integrated new classifiers designed to intercept and block cybersecurity-related tasks. Commerce Secretary Howard Lutnick confirmed that government officials worked closely with the company over the past two weeks to ensure the models met national security requirements. Moving forward, the firm plans to deepen its collaboration with the U.S. government, including providing pre-release access for safety evaluations and participating in the Project Glasswing framework to standardize how the industry assesses the severity of AI jailbreaks.