France Flags AI Lock-In Risk as Three Companies Control 84% of the Market
The French competition authority's report is a warning shot. The question is whether anyone can do anything about it.
Published: 22 July 2026 Category: Policy / Competition Sources: PPC Land
The Number
France's Autorité de la Concurrence published a report this week with a striking statistic: OpenAI, Google, and Anthropic together control 84% of the global AI agent market. The figure is based on enterprise deployment data, API call volume, and revenue estimates. It is not a precise measurement — the AI agent market is fluid and poorly defined — but the direction is clear. The market is concentrating rapidly, and the concentration is happening around American companies.
The Argument
The French authority's concern is not just market share. It is lock-in. Enterprises that build workflows around ChatGPT Work, Claude Cowork, or Google's agent tools are making investments — training employees, integrating systems, customising prompts — that become sunk costs. Switching to a competitor becomes expensive and disruptive. The result is sticky customers and limited competition, even if better alternatives exist.
The report recommends several interventions: interoperability requirements, data portability mandates, and restrictions on exclusive cloud compute deals. These are standard competition tools, and they are standard because they often fail. Interoperability requirements sound good in theory but are nightmareish to implement in practice. Data portability assumes that the valuable part of an AI deployment is the data, when increasingly it is the fine-tuned models, custom tools, and institutional knowledge embedded in prompts and workflows.
The Analysis
The deeper problem is that AI agent markets may be naturally concentrated. The fixed costs of training frontier models are enormous. The returns to scale are significant — more users means more feedback means better models. The network effects are real — an agent that integrates with more tools is more useful than one that integrates with fewer. These are characteristics of markets that tend toward oligopoly, and regulation can nudge the outcome but may not be able to prevent it.
France's report arrives at a politically complicated moment. The EU is already implementing the AI Act. The US is in a protectionist posture, unlikely to support measures that disadvantage American companies. China is producing competitive models but is not part of the western regulatory conversation. Global coordination on AI competition policy is not on the horizon.
The Verdict
The 84% figure is useful as a rhetorical device. It captures attention. It frames the issue as urgent. But the policy response is uncertain, and the market dynamics that produced the concentration are powerful enough that modest regulatory interventions may not change the trajectory.
What could change it? A genuinely competitive open-source ecosystem would help. So would public investment in non-commercial AI infrastructure. So would antitrust enforcement with teeth — blocking acquisitions, unwinding exclusive deals, requiring model sharing as a condition of market access.
None of these are likely in the current political environment. France will issue its report. The EU may draft some regulations. The US will do nothing. The market will continue to concentrate. And the 84% figure, a year from now, will likely be higher.
The warning shot has been fired. Whether anyone is listening remains to be seen.