Commerce Secretary Lutnick Sent Formal Letter Threatening Anthropic with Criminal Penalties Over Fable 5 Access

This is a developing story. NCT previously covered the initial export control directive, Anthropic’s negotiations with Washington, and the role of Amazon CEO Andy Jassy in raising concerns.

Commerce Secretary Howard Lutnick sent a formal letter to Anthropic last week threatening criminal and civil penalties if the company grants foreign nationals access to its most advanced AI models without government permission, Bloomberg reported on June 16. The letter marks the first documented legal threat in the Fable 5 and Mythos 5 export control escalation, moving beyond the verbal directive issued on June 13.

From Verbal to Written

The Trump administration’s original export control directive was communicated verbally. Anthropic said at the time it had received only “verbal evidence of a potential narrow, non-universal jailbreak.” The company disabled both Fable 5 and Mythos 5 for all users, including US customers, because it could not verify citizenship at the API level.

Lutnick’s letter formalizes the restriction. According to Bloomberg, it explicitly states that Anthropic must obtain government permission before granting foreign nationals access to its advanced models and warns of criminal and civil penalties for non-compliance. This transforms the directive from an ambiguous verbal order into an enforceable legal threat with specified consequences.

The Security Debate Intensifies

The underlying trigger for the export controls remains contested. Katie Moussouris, founder of Luta Security and the only outside expert to review the third-party research paper that prompted the ban, wrote on Monday that the reported “jailbreak” was actually a simple prompt: “fix this code.” Researchers fed Fable 5 open-source code with known CVEs and asked the model to fix them. It complied.

“‘Fix this code,’ plus several manual steps to generate test scripts, should never have triggered an export control,” Moussouris wrote, according to The Register. More than 100 cybersecurity leaders signed an open letter urging the administration to reverse the restrictions, arguing that removing the best AI tools from defenders benefits attackers.

The EU Response

The European Commission is using the situation to advance its case for technological sovereignty. Thomas Regnier, spokesperson for the Commission, told The Register that Brussels is “assessing its implications, including for users in the European Union” and characterized it as evidence that Europe must achieve AI independence. Chinese rivals including Zhipu, whose stock surged 33% this month partly on the Anthropic restrictions, stand to benefit from any sustained access gap.

The Enforcement Question

Lutnick’s letter raises a practical question for the broader AI industry: if the Commerce Department can issue export control directives on AI models based on a single research report about defensive security prompts, what prevents similar restrictions on other frontier models? OpenAI, Google DeepMind, and Meta all ship models with comparable or greater cybersecurity capabilities. None have received similar letters.

Anthropic executives visited Washington, D.C. last week to negotiate a resolution. Whether the formal letter represents a hardened position or a prerequisite for structured negotiations remains unclear. Both models remain offline for all users.