Anthropic has announced that global access to its most advanced AI models, Claude Fable 5 and Claude Mythos 5, has been restored after export restrictions imposed by the U.S. Department of Commerce were lifted. The company said it received a notice from the Commerce Department and began implementing the lift of restrictions immediately.
Only four days earlier, Anthropic had reported receiving government authorization to provide access to Mythos 5 for a narrow U.S. cybersecurity cohort. The two models had been suspended by U.S. authorities on June 12 after vulnerabilities were found in their built‑in safety systems that could not fully rule out potential misuse of the technology.
The export ban appears to have been removed because U.S. officials concluded that Anthropic had taken appropriate steps to mitigate the risks associated with Claude Mythos 5 and Claude Fable 5.
Background and government tensions
Earlier reporting noted that as part of a security test an Anthropic model penetrated systems at the U.S. National Security Agency (NSA); that incident happened one day before the Trump administration ordered the suspension of access for foreign nationals on national‑security grounds.
Relations between Anthropic and some U.S. government agencies had been strained even before the export controls. The company refused to permit the U.S. military to use its models for domestic surveillance and fully autonomous weapon systems, and the Pentagon subsequently placed Anthropic on an official blacklist. Despite that, the NSA has reportedly continued to use the company's AI. Anthropic has said it will pursue legal action against the sanctions.
Why this matters
Access to high‑capacity language models is a sensitive issue with geopolitical and national‑security implications. The Commerce Department's decision to lift restrictions suggests authorities are satisfied with the mitigation measures Anthropic implemented, but earlier incidents and ongoing tensions underscore continuing questions about oversight and acceptable uses of advanced AI.
