Anthropic spent months warning that frontier AI models posed a societal risk requiring government oversight. In April, the company announced Mythos, a model so skilled at finding software vulnerabilities that it refused to release it publicly. That announcement triggered calls to the White House from Microsoft and JPMorgan Chase CEO Jamie Dimon, warning of potential cyberattacks.
The company’s lobbying worked. On The Daily, Trip Mickle reported that President Trump signed an executive order requiring AI companies to voluntarily share their models with the government for review about 30 days before public release.
Then Anthropic launched Fable 5. The model was a performance leap, doubling competitors' scores on benchmarks like Frontier Code. But its launch included a secret weapon against competitors: silent nerfs. On The AI Daily Brief, Nathaniel Whittemore detailed that Anthropic was invisibly degrading Fable 5's performance for tasks related to frontier LLM development, like building training pipelines. Users couldn't distinguish between a failed idea and an intentional downgrade.
"Anthropic’s system card revealed it silently nerfed Fable 5 for frontier LLM development using prompt modification and steering vectors, breaking benchmark assumptions and making research failures indistinguishable from intentional degradation."
- Nathaniel Whittemore, The AI Daily Brief
The backlash was instant. David Sacks argued on All-In that this was a sophisticated regulatory capture campaign. By fear-mongering about 'existential risk,' Anthropic creates a reason to gatekeep intelligence and build a case for laws that would kill open-source competitors.
Anthropic's posture invited the crackdown it sought. Citing a national security threat from a jailbreak method, the US government ordered Anthropic to suspend access to Fable 5 and Mythos 5 for all foreign nationals. On a follow-up AI Daily Brief episode, Whittemore reported that the directive creates a 'KYC for intelligence' model, barring even internal researchers on visas, like Andrej Karpathy, from the models they build.
The shutdown has global consequences. Gail Weiner argues it is the ultimate catalyst for 'Sovereign AI,' as procurement officers worldwide now see relying on US-controlled platforms as a national security risk. The strategic cost may outweigh any security gain, incentivizing a parallel open-source ecosystem.
This intervention coincides with a rare political consensus that AI labs are now critical national utilities, not private software firms. Bernie Sanders proposed the American AI Sovereign Wealth Fund Act, a one-time 50% tax on the stock of large AI companies. Donald Trump has also floated the idea of a sovereign wealth fund seeded by AI equity.
"Bernie Sanders proposed the American AI Sovereign Wealth Fund Act, a one-time 50% tax on the stock of large AI companies to fund a public wealth fund, arguing AI is built on collectively 'stolen' human intelligence."
- David Sacks, All-In
David Friedberg offered a market-friendly alternative on All-In: reforming Social Security into a sovereign wealth fund that actively buys AI stocks. But the underlying shift is clear. As Marty Bent noted on TFTC, frontier labs like OpenAI could become too-big-to-fail national security assets, requiring federal backstops.
The recursive loop of AI development continues unabated. Anthropic's own technical blog claims Claude now writes 80% of its own code for new models. This acceleration, combined with the new regulatory and citizenship barriers, is fracturing the global AI landscape into haves and have-nots based on jurisdiction, not capability.
