A theory about pricing, regulation, and who gets to use AI

I have a theory.

Peter Thiel's Founders Fund co-led Anthropic's $20 billion raise in February. Thiel is also a co-founder of Palantir, whose relationship with government and regulation is well documented. His view is that technology can reshape and overcome the constraints imposed by bureaucracy.

The subscription versus enterprise pricing model at Anthropic is interesting through that lens. The $100 Pro subscription is viable for individuals. Enterprise pricing, with the audit trails, data governance, and security controls that regulated industries require, is a different order of magnitude. I spent $4,000 in enterprise tokens in one week just working. Not trying to. Just working.

The economic gap between what an individual can use and what a compliant enterprise can afford creates a forcing function. Organisations that cannot absorb the enterprise cost either run without the controls, find a balance, or build their own models. The policies and regulations that require those controls, GDPR, the EU AI Act, sector-specific requirements, become economic liabilities rather than governance tools.

Whether intentional or not, the effect is to make regulatory compliance expensive and to reward those who can either afford it or ignore it.

The Fable shutdown made this starker. The U.S. government ordered Anthropic to disable Fable globally last Friday, with a few hours notice, citing national security. Every company building on those models lost access instantly. Anthropic complied but made clear it disagreed with the order.

The open weight escape route, run your own model and control your own access, becomes the logical response. Except the capability gap between open weight models and frontier models is widening. If that gap keeps expanding, opting out of the commercial model means accepting a capability disadvantage you may not recover from.

So the question is not just about pricing. It is about who controls access to the tools that are becoming infrastructure. And whether the governance frameworks being built around those tools are designed for the organisations that can afford them, or for everyone.