Same Gatekeepers, New Gateway: Open Markets warns Google, Amazon, Microsoft and Meta could control AI agents

The argument is not that Big Tech already runs AI agents. It is that the same four companies that run today’s gateways — cloud, models, operating systems, browsers, search, social platforms — are the ones best placed to run the next gateway too. A search engine influences what you see and choose, the report says, but still leaves the choice to you. An agent chooses for you. That is the difference the whole report turns on.

Force one: who gets distributed#

Distribution is the oldest trick in the platform book, and the report says it transfers cleanly to agents. If your agent ships by default on the phone, in the browser, or inside the messaging app billions of people already use, rivals have to fight for every install while yours arrives pre-installed. The report points to Europe, where regulators have already stepped in to guarantee agents access to Google’s Android ecosystem and Meta’s WhatsApp — disputes it reads as early proof that gateway owners can suppress competition in the agent market before it matures.

“Big Tech already controls much of the infrastructure that powers the internet. Now those same companies want to control the AI agents that will increasingly decide what people buy, read, and do,” said Max von Thun, Director of Open Markets Institute Europe. “If policymakers fail to act, we risk transforming today’s digital gatekeepers with something even more dangerous: corporations that make choices for us.”

Force two: the data you can’t take with you#

The second force is quieter and harder to regulate away. An agent that has handled your email, your calendar, your shopping and your messages for a year knows things a new agent does not — your preferences, your history, your habits. Starting over means retraining a stranger. The longer you stay, the more expensive leaving gets, and the early leaders keep the lead.

The concentration is already measurable. In an advisory opinion dated July 17, 2026, France’s competition authority found that OpenAI, Google and Anthropic held over 84% of the global AI agent sector in May 2026, citing Sensor Tower data, and flagged default integration, data advantages and switching difficulty as lock-in risks. Dr. Courtney C. Radsch, the report’s co-author and Director of the Center for Media and Digital Governance, puts the endgame bluntly: “As agentic systems grow more sophisticated and concentrated there is a serious risk that predictive personalization will become anticipatory persuasion.” Agents that act on your behalf are already here in narrow forms — one now drives a real iPhone — and every task they complete adds to the profile that makes them hard to leave.

Force three: who sets the rules#

The third force is about the plumbing. Agents depend on infrastructure, protocols and services owned by someone — and whoever owns them sets the terms of participation. The report’s examples are already visible in social feeds: xAI embedded Grok as a dedicated tab inside X, and Meta is experimenting with a similar move in Threads. When the agent lives inside the platform, the platform decides what the agent can reach, what it can sell, and whose content it surfaces.

This is also where the report’s asks get sharpest. It wants enforceable duties of loyalty and care — a legal requirement that agents serve their users rather than their owners’ commercial interests — and it wants technical standards set through transparent, representative governance instead of by the dominant corporations themselves. The final ask is the bluntest: antitrust enforcement to stop anti-competitive practices, block or unwind anticompetitive mergers and partnerships, and pursue structural remedies where vertical integration creates conflicts of interest.

The verdict#

The report’s strongest claim is its timing argument: unlike search or social media, the agent market is still taking shape, so regulators have a window to act before the rules are written by the companies that benefit from them. European competition authorities are already addressing interoperability and access barriers for AI services, and U.S. lawmakers have begun considering legislation on duties and safeguards for agents. The report argues those efforts need to become one strategy covering the whole stack.

The honest caveat is that the report is an advocacy document, not a measurement exercise. It presents no new data of its own; its evidence comes from news coverage, company announcements, academic papers and third-party estimates. PPC Land, which covered the launch, flags two errors in it — figures for Anthropic that differ between pages 28 and 54, and a February 2026 date for Instant Checkout that does not match OpenAI’s September 2025 announcement. None of the companies the report criticizes is quoted responding to it. The argument stands or falls on its logic, not on new evidence — and the logic, that whoever controls distribution, data and standards controls the agent, is hard to dismiss even if the footnotes wobble.

What to watch#

Whether European regulators extend the Android and WhatsApp access interventions into a general rule for agent access, and whether U.S. lawmakers turn the duties-of-loyalty proposals into actual legislation. Whether the four named companies answer the report at all — silence so far — and whether Meta’s own agent push and the enterprise agent land-grab make the concentration numbers move before any rule lands. If the next year brings one default agent per platform and no portability rule, the report’s warning will have been proved right by the market it was trying to warn about.

Sources: Open Markets Institute, “Same Gatekeepers, New Gateway: AI Agents, Monopoly Power, and Democratic Risk” (November 9, 2026), here; PPC Land, “Google, Amazon, Microsoft, Meta could dominate AI agents, Open Markets warns” (November 9, 2026), here.