The Fable/Mythos Precedent: Export Control as Discretionary Instrument in U.S. Frontier AI Policy
Authors/Creators
Description
Working paper, v2 (June 2026). Methodology: FACT (verifiable on primary source) /
INFERENCE (logical deduction from documented facts, falsifiable). Speculation excluded.
On June 12, 2026, the U.S. Department of Commerce ordered Anthropic to suspend global
access to Claude Fable 5 and Mythos 5 for any foreign national, with ninety minutes'
notice. Fourteen days later, OpenAI's comparably capable GPT-5.6 family entered a
staggered, partner-vetted preview under a different and far less disruptive legal
instrument. This paper documents the legal architecture of the Anthropic directive,
the contested and partially withdrawn technical justifications offered for it, the
absence of a declared capability threshold separating the two cases, the cabinet-level
rivalries that managed the dispute in place of a coherent regulatory process, and the
extraterritorial consequences visible in France and Germany. It also documents two
claims that circulated as evidence in this case - a Senate-relayed account of an NSA
"breach" and a UK minister's resignation - that this paper's own verification process
found unsupported or contradicted by primary sources, and explicitly withdraws them
(Section 6).
Conflict of interest: this paper was produced using Anthropic Claude as a research
and drafting tool. Anthropic and its Fable 5/Mythos 5 models are the central subject
of the analysis. Independent adversarial review on a non-Anthropic model was conducted
prior to this v2; a residual conflict-of-interest risk on the comparative
Anthropic/OpenAI framing (Sections 3-4) is disclosed and not fully resolved. Full
statement in the document.
v2 changelog: removes an unverified exact EAR statutory citation and an unverified
Financial Times attribution; downgrades a bipartisan House letter claim to unverified
aggregated reporting; explicitly labels all single-source claims; de-quotes IMF/ECB
material pending transcript verification; reframes Section 4's central claim from
causal to descriptive; adds a verification-status table (Section 7); strengthens the
conflict-of-interest statement.