A 3,700-page report from France’s competition authority raises new questions about who controls the future of AI assistants
France’s competition authority has published an extensive advisory opinion warning that the market for AI agents, the increasingly autonomous software tools capable of completing tasks on a user’s behalf, has become highly concentrated in the hands of just three companies. According to the report, OpenAI, Google and Anthropic together control more than 84% of this rapidly growing market. For everyday consumers who may not think much about which company built the AI assistant helping them shop, plan travel or manage daily tasks, the finding raises a quieter but important question: does it matter who controls the technology increasingly making decisions on our behalf?
How Regulators Investigated the Market
What sets this investigation apart from a typical regulatory report is the hands-on approach French authorities took to reach their conclusions. Rather than relying solely on written submissions and interviews, regulators built and deployed their own AI agents, then ran them through 550 separate shopping-related prompts to observe firsthand how these systems behave in real-world conditions. The inquiry, which began in January 2026 and concluded with a report spanning roughly 3,700 pages, represents one of the more thorough regulatory examinations of AI agent behavior conducted anywhere to date, and stands out for testing actual agent performance rather than relying purely on company disclosures and public statements.
The investigation’s findings describe a market that has shifted meaningfully in a short period of time, moving from simple chatbots capable of answering questions to autonomous agents that can reason through multi-step problems and coordinate real-world actions, such as completing purchases or booking services, largely without direct human oversight at each step. This evolution from passive question-answering tools to active decision-making systems is precisely what has drawn the attention of competition regulators, since a market dominated by so few companies raises the stakes considerably when those companies’ products are making purchasing decisions or other consequential choices for millions of users.
Why Market Concentration Matters for Ordinary Users
Market concentration in any industry can lead to reduced competition, higher prices and fewer choices for consumers, but the concern takes on a different dimension when the products in question are AI agents actively making decisions rather than simply displaying information. If the vast majority of AI agent interactions flow through just three companies’ underlying technology, those companies gain significant influence over which products get recommended, which services get booked, and ultimately which businesses benefit from AI-driven consumer decisions, an influence that smaller companies and startups may struggle to counterbalance without comparable resources.
This dynamic mirrors concerns regulators have raised in the past about search engines and app stores, where a small number of dominant platforms effectively serve as gatekeepers between consumers and the broader marketplace. The difference with AI agents is that the gatekeeping function is less visible to the average user, since an AI agent completing a task on someone’s behalf may not clearly disclose why it recommended one option over another, making this particular form of market concentration potentially harder for consumers to recognize or evaluate compared with more familiar examples like a dominant search engine or app marketplace.
What This Means for Smaller Companies and Startups
For startups and smaller technology companies trying to compete in the AI agent space, this level of concentration presents a genuine structural challenge. Building a competitive AI agent requires enormous computing resources, access to large amounts of training data, and the kind of sustained capital investment that only a handful of companies can currently sustain, which helps explain why the market has consolidated so quickly around the same three names identified in the French report. Smaller companies attempting to enter this space often find themselves building on top of the very infrastructure provided by these dominant players, which can limit how independent their own products can truly become, even when marketed under a different brand name.
This same dynamic has begun drawing attention from regulators beyond France as well. The European Commission has recently moved to formalize updated timelines under its broader AI Act framework, alongside new guidance on transparency requirements for AI-generated content, signs that European regulators more broadly are trying to build oversight structures capable of keeping pace with how quickly this market has consolidated around a small number of dominant firms.
What Happens Next
France’s competition authority has issued this document as an advisory opinion rather than a binding enforcement action, meaning immediate changes to how these companies operate are unlikely in the short term. Advisory opinions of this kind typically serve as a foundation for future regulatory action, both within France and potentially across the broader European Union, where competition regulators have shown increasing willingness to scrutinize dominant technology companies in recent years. The scale and detail of this particular report, running to thousands of pages and grounded in direct testing rather than solely theoretical analysis, suggests French authorities intend for it to carry significant weight in future discussions about how AI agent markets should be regulated going forward.
For consumers, the practical takeaway for now is less about any immediate change and more about awareness. As AI agents become more common in everyday tasks, understanding that a small handful of companies control the vast majority of the underlying technology may be worth keeping in mind when deciding how much autonomy to hand over to these tools, particularly for decisions involving money or personal information.
Sources consulted:
Ink & Algorithms, AI Weekly Pulse: https://inkandalgorithms.systeme.io/ai-weekly-pulse-4-ai-model-releases-july-2026
Stephenson Harwood, Neural Network: https://www.stephensonharwood.com/insights/neural-network-july-2026/
