Regulation is a Proven Way of Protecting Industry Incumbents

Are industry leader calls for regulation of artificial intelligence models a tacit form of market leader moat building, as has proven to be the case for other heavily-regulated industries?

The basic economic mechanism is well established: regulation creates fixed costs, and fixed costs favor firms with scale. Regulation also limits and shapes key contestant options in ways that can protect them from excessive and violent spending or price competition.

The analogy might be rate regulation in any capital-intensive industry such as passenger airlines, electricity or natural gas supply or telecommunications.  

In Lewis Carroll’s Through the Looking Glass, the Red Queen tells Alice: “Now here, you see, it takes all the running you can do, to keep in the same place. If you want to get somewhere else, you must run twice as fast as that!”

And that’s the analogy to the pace of frontier model development: contestants are forced to spend sums financial observers do not like, just to stay where they are, competitively. You cannot afford to stop investing because your competitors are investing.

And raising fixed costs in the form of regulation makes it harder for upstarts to catch up. So today’s market leaders get some spending relief while challengers face higher costs if they want to try and catch up. 

Brookings made essentially this point specifically about AI licensing, noting that licensing can reinforce the position of companies that already possess the necessary approvals. An academic analysis of foundation-model competition likewise warned that the early calls for regulation were coming from major industry participants and could raise entry barriers. 

What the industry says	Potential strategic effect for an incumbent
“Frontier AI is too dangerous to develop without government oversight.”	Might well favor closed model developers more than open source developers, which are a key market challenger.
“Companies above a capability threshold should be licensed.”	Turns massive compute, safety teams and regulatory expertise into barriers to entry.
“Models should undergo expensive third-party testing.”	Adds fixed costs that a well-funded incumbent can absorb more easily than a startup.
“Developers must maintain sophisticated security and monitoring systems.”	Favors companies that already have large compliance and security organizations.
“We need standardized safety evaluations before deployment.”	Slows the frequency with which new models can be released and potentially reduces price competition.
“The industry should coordinate on safety standards.”	Can reduce the incentive for an incumbent to engage in an expensive capability race—although coordination among competitors raises antitrust issues.
“Government should establish clear rules rather than allowing a patchwork of state laws.”	A national framework can be easier for large incumbents to navigate than 50 different regimes. OpenAI has explicitly advocated harmonized national standards for this reason. (OpenAI)
“Only the most capable systems need stringent controls.”	Lessens market leader competitive threats by subjecting all to constraints.

The “regulatory moat” strategy would make it expensive and difficult for anybody else to become a frontier competitor.

The preferred rules would therefore tend to emphasize:

• licensing

• minimum compute/security requirements

• mandatory evaluations

• expensive certification

• reporting

• government audits

• liability restrictions on open-weight models

• controls on access to advanced chips.

An  “arms-race brake” strategy also applies. The objective isn't to eliminate competitors but to make the pace of competition more manageable.

Investment demands slow If every frontier company knows that once a model crosses a particular capability threshold it must:

• undergo independent evaluation

• demonstrate cybersecurity

• document dangerous capabilities

• establish monitoring

• satisfy government requirements

• perhaps obtain approval before deployment.

That should, in theory, lower the financial returns for the next N unit of performance enhancement, which is what frontier-model developers already are seeing. 

But that all works only when regulation slows all contestants equally. If only firms in one country agree to the regulation, while those in other countries do not have to comply, the regime doesn’t work.

From a leading supplier perspective, there is an incentive to support a regulatory regime that raises the fixed cost of frontier development and slows the competitive capability race, provided that the regulatory burden falls disproportionately on potential entrants and foreign competitors rather than on themselves.

If that can be accomplished, market leaders gain:

• a competitive moat

• more capital investment discipline

• Public policy legitimacy (“we are promoting safety, not protecting our market leadership”)

A company doesn't have to fabricate AI risks for regulation to serve its competitive interests.