Mythos AI, Anthropic’s classified research model, cracked nearly every NSA system in hours - not weeks - according to Senator Mark Warner, citing Cyber Command chief General Joshua Brudd. This breach triggered Trump’s decision to block the commercial release of Fable 5, the model’s public-facing version.
The move wasn’t driven by independent oversight. According to Krystal Ball on Breaking Points, Amazon - Anthropic’s chief rival through its $50 billion investment in OpenAI - was the entity that flagged Fable 5 as a national security threat. The timing reeks of competitive manipulation: a dominant player leveraging state power to delay a competitor’s product under the guise of safety.
"Mythos AI broke into almost all U.S. classified systems not in weeks, but in hours."
- Senator Mark Warner, Breaking Points with Krystal and Saagar
On Rabbit Hole Recap, Odell argued the safety narrative is a Trojan horse for regulatory capture. By promoting the idea that frontier AI is inherently dangerous, firms like Anthropic invite government intervention that creates moats against open-source and smaller competitors. But in this case, the irony is thick: the government didn’t act on Anthropic’s own warnings - it acted on a tip from Amazon.
Six days after the ban, the lack of a formal regulatory framework is glaring. Saagar Enjeti notes the Trump administration’s AI policy is ad hoc, driven by CEO politics rather than consistent standards. If one company can kill another’s product via backchannel warnings, the system isn’t about security - it’s about control.
"This isn't a safety failure. It's a calculated bid for regulatory capture."
- Odell, Rabbit Hole Recap
The precedent is dangerous. A high-performing AI model is pulled not through judicial or legislative process, but by executive assertion backed by a competitor’s complaint. The national security justification may be valid - but when the accuser profits from the outcome, trust evaporates. The AI race is no longer just technological. It’s political.


