The US government is restricting Anthropic's commercial rollout of its new model, Mythos, over concerns it could hamper the government's own access to compute. This move treats AI capacity as a strategic national resource and effectively creates a de facto licensing system for powerful models, marking a new era of AI governance.

The narrative of AI doom isn't just organic panic. It's being leveraged by established players who are actively seeking "regulatory capture." They aim to create a cartel that chokes off innovation from startups right from the start.

Bill Gurley voices concern that large AI companies like Anthropic, which are lobbying heavily, might be using regulation as a competitive weapon. This "regulatory capture" tactic would create high barriers to entry, stifling innovation from smaller startups and open-source projects, effectively "pulling up the ladder" behind them.

As enterprises replace expensive proprietary models with cheaper open-source alternatives, frontier labs like OpenAI and Anthropic face an existential threat. Their strategic response could be to lobby for regulations that effectively make open-source models illegal, creating a protective moat.

Andreessen recounts meetings where officials detailed a plan to control AI by limiting it to 'two or three big companies working closely with the government.' This strategy involves protecting these giants from startup competition and even classifying the underlying math to centralize power.

Leading AI companies allegedly stoke fears of existential risk not for safety, but as a deliberate strategy to achieve regulatory capture. By promoting scary narratives, they advocate for complex pre-approval systems that would create insurmountable barriers for new startups, cementing their own market dominance.

Silicon Valley's economic engine is "permissionless innovation"—the freedom to build without prior government approval. Proposed AI regulations requiring pre-approval for new models would dismantle this foundation, favoring large incumbents with lobbying power and stifling the startup ecosystem.

Slowing public releases of AI models for government review may not slow overall progress. This creates a scenario where labs advance internally for months, giving government agencies exclusive access while delaying public commercialization and the next cycle of investment.

The fear of killer AI is misplaced. The more pressing danger is that a few large companies will use regulation to create a cartel, stifling innovation and competition—a historical pattern seen in major US industries like defense and banking.