The Once And Future Fable #4
Zvi's AI Roundups · Zvi Mowshowitz · 2026-06-24
Zvi Mowshowitz analyzes the US government's export control restricting a frontier AI model referred to as 'Fable 5,' arguing the measure was a poorly executed policy fiasco that damaged allies' trust, was triggered by misunderstood NSA red-team results, and will not slow AI development.
Appears in
Extraction
Topics: ai-policyexport-controlscybersecurityfrontier-modelsus-china-ai
Claims
- The US export control taking 'Fable 5' offline was a policy fiasco that damaged the credibility of the American AI stack among allies and caused the NSA itself to lose access to the model.
- NSA red-team results claiming the model 'broke into classified systems in hours' were widely misunderstood—the tests involved authorized analysts with physical access to air-gapped systems, not an outside attacker scenario.
- A newer, more capable model version has already completed training, demonstrating that restricting public access does not slow underlying AI development and may slightly accelerate it by freeing resources.
- Meta is the last major AI lab holding out on voluntary US government model review, while OpenAI, Anthropic, Google, xAI, and Microsoft have all agreed.
- US-China talks on AI guardrails fell short because China sent foreign ministry officials rather than technical experts, and the US accepted this arrangement.
Key quotes
Even if access is restored in the coming week as expected, this was a fiasco.
Stopping models like Fable 5 or Mythos 5 from being served to the public does nothing to slow down development. In fact, it probably speeds it up slightly by freeing up resources.
The chosen implementation was such a train wreck that not only did we piss off our allies and call into question the reliability of the entire 'American AI stack,' we also caused the NSA to lose access to Mythos.