Home

Donate
Perspective

The Next Big Tech Monopoly Could Be Your Personal AI Agent

George Colville / Nov 20, 2026

Joudy Bourghli & The Bigger Picture / Better Images of AI

Republish

Muse, Dots, Duet, Spark, Copilot, Grok Bot. Agentic AI is now a mass-market consumer product.

In tech hubs like San Francisco, New York, and London, expert users have been employing agents, both individually and in swarms, on work tasks and side projects. They have used agents to build personal apps, run cafes, book restaurants and tennis courts, evaluate health insurance plans, and register with new doctors. Until now, this was the prerogative of the tech-savvy. That is about to change.

This is no bad thing. The prospect of democratizing the time- and labor-saving benefits of agentic AI beyond expert users is exciting. That promise, however, comes with potential peril. Beyond the cybersecurity and privacy risks, and the societal upheaval as systems built for a time before AI strain under its pressure, lies a deeper, systemic threat. This is the possibility that an agentic AI monopoly or oligopoly concentrates extraordinary power in the hands of a small number of private corporations. In a new report, “New Gateway, Same Gatekeepers,” we examine this threat, identify the economic and democratic risks it poses to society, and set out what we can do to avert it.

Personal agents: the endgame for consumer AI

While the high priests of AI continue to prophesy artificial general intelligence, the industry built around them has been hard at work designing products it can sell. Unlike businesses, which have been experimenting with agentic AI for months or even years, consumer interactions with AI have generally been limited to conversations with chatbots. That changed a few months ago. Since mid-2026, all the major AI players – OpenAI, Anthropic, Meta, Google, Microsoft, and xAI – have released new accessible personal AI agents that can act on people’s behalf.

The ambition for these companies is simple: to create and monetize a personal assistant that can automate as many daily tasks as possible. They want, in other words, to become a universal digital intermediary: a single platform through which people buy their groceries and their clothes, book their holidays, interact with government services, control their homes, consume news, and communicate with their friends, family, and communities. Instead of dozens of apps and websites, the intention is for one single personal assistant with a text or voice interface to be the only interface people interact with for most of their digital needs.

This vision is compelling. It could free people up to spend more time with their friends and families, on their hobbies, and on their work. At the same time, it raises deeply troubling questions for individual autonomy, economic fairness, and democracy.

New digital gatekeepers with unprecedented power

While the promise of personal agents is vast, there is a real risk that one or a small number of companies come to dominate the technology, establishing an agentic AI monopoly or oligopoly. In doing so, these corporations would become the arbiters of what people buy and who they buy it from, would select which news they receive, and would shape how they communicate with their friends, family, and communities.

They could do this fairly, but commercial pressure and the lack of competitive incentives to treat customers well will almost certainly push them in more harmful directions. With access to vast quantities of personal data, and acting on users’ behalf as a kind of digital exoskeleton, they would open the door to new extremes of surveillance and exploitation, both political and commercial.

A concentrated market would create gatekeeper dynamics even more extreme than Amazon’s, Google’s and Apple’s existing versions. An agentic AI gatekeeper could charge excessive commissions to merchants that sell their products to people via AI agents. It might prioritize answers and actions that favor partners and advertisers, turning the entire economy into a pay-to-play system. Extreme forms of surveillance pricing would also become possible as monopolists gain access to the unprecedented quantities of personal data funneled through agents.

Beyond this, concentrated control of agentic AI would vest incredible political power in the hands of private actors. People already increasingly rely on AI agents and assistants to access news and information, including on political issues, and to make decisions based on that information. Whoever controls these tools shapes what information people see and how it is presented to them, creating opportunities for both direct and indirect influence.

In an AI oligopoly, that control rests in the hands of a small number of private firms or states with influence over them. With agentic AI, political power goes beyond information towards potential coercion. The corporations controlling agents may end up holding the reins to significant parts of our economies and our access to essential services: power it can instrumentalize to coercive political ends.

The road to concentrated agentic AI

Agentic AI could give a small number of private actors unprecedented power over vast swathes of the global population, power they are likely to misuse and abuse. In a competitive market, where users can easily change agent, these risks are lower: competitive pressure helps moderate abusive behavior by allowing users to jump ship to a different provider if they are badly treated. However, if one or a few firms come to dominate and people become locked into using a small number of agents, the risks rapidly become daunting.

Instead of a market characterized by perfect competition, personal agents are being introduced into a world of deeply embedded tech monopolies. These giants control digital empires spanning hardware, operating systems, browsers, search engines, email, messaging services, social media, productivity software, and online marketplaces. They enter the fray with enormous capital resources and the ability to distribute and connect their agents in ways nobody else can compete with.

While Apple continues its long struggle to make Siri workable, Google, Meta, and Microsoft have each released personal agents which leverage their existing dominant power in other areas.

AI and platform power

How five tech giants tie their AI to the rest of their products

Each company’s AI assistant draws on products it already owns, and many of those products steer users toward the company’s own services. Select a company to see its products.

Self-preferenced distributionProprietary integration and/or data sharingBothNeither
GoogleAI assistant:Gemini “Personal Intelligence”29 products

Hardware

Pixel phoneGooglebook

Browsers & operating systems

AndroidChromeGoogle TV OSChromeOS

Productivity

DocsSheetsSlidesDriveKeepCalendarTasksTranslate

Communications

GmailMeetChatContacts

Search & discovery

SearchMapsNews

Marketplaces & e-commerce

ShoppingFlightsHotelsGoogle Pay

Media & entertainment

YouTubePhotosPlayTV/Chromecast
MicrosoftAI assistant:Copilot18 products

Hardware

SurfaceXbox

Browsers & operating systems

WindowsEdgeXbox OS

Productivity

WordExcelPowerPointOneNoteOutlook CalendarOneDrive

Communications

OutlookTeams

Search & discovery

BingMSN

Software development

GitHubVS Code

Social & professional networking

LinkedIn
MetaAI assistant:Muse10 products

Hardware

AI glasses

Social networking

FacebookInstagramThreads

Communications

WhatsAppMessenger

Marketplaces & e-commerce

Facebook MarketplaceFacebook ShopsInstagram ShoppingMeta Pay
AppleAI assistant:Siri / Apple Intelligence29 products

Hardware

MacBook laptopsDesktop PCsiPhoneiPadApple WatchAirPodsHomePodApple TV

Browsers & operating systems

iOSmacOSwatchOStvOSSafari

Productivity

PagesNumbersKeynoteNotesMailCalendariCloud Drive

Entertainment

MusicTVNewsFitnessPhotos

Communications

MessagesiCloud Mail

Commerce

Apple Pay

Software development

Xcode
xAIAI assistant:Grok1 product

Social networking

X (formerly Twitter)
Source: Open Markets Institute Europe, “Same Gatekeepers, New Gateway: AI Agents, Monopoly Power, and Democratic Risk,” Nov. 9, 2026. openmarketsinstitute.eu
Graphic: Tech Policy Press

The tech giants’ first advantage is in distribution. Their existing platforms allow them to instantly put their AI products in front of millions – even billions – of users. Meta’s Muse, as well as having its own app, is directly integrated into WhatsApp with its 3 billion monthly users. While Google’s Spark agent is currently only available through Gemini apps, Google has been distributing its AI services through Search, Chrome, Android, and its Workplace suite, and will almost certainly do the same for its agents. Microsoft, meanwhile, is boosting its Copilot assistant via its Microsoft 365 productivity apps, GitHub, and the Windows operating system.

Integration deepens the advantage. Meta’s Muse has privileged access to Facebook, Instagram, and Facebook Marketplace on top of WhatsApp. Google’s Spark has unmatched integration with its Workspace, and its “Personal Intelligence” spans Gmail, Maps, Flights, Shopping, and YouTube. Microsoft’s Copilot connects to Outlook, Teams, and the wider Office suite. Rivals cannot match this access.

Beyond these uses of existing monopoly power to gain unfair advantages, the nature of agentic AI technology itself may also concentrate power. Our analysis identifies two mechanisms working in favor of this outcome.

First, network effects: if agents are not interoperable, the more developers and service providers build for a given agent, the more useful it becomes to users, attracting more users and thereby making that agent the first platform developers and service providers build for. Rivals are left with a thinner ecosystem and a weaker case for users to switch.

The second is data feedback loops: an agent that learns from its interactions improves with use, so the provider with the most users accumulates the most data and builds the best agent, which attracts still more users.

Additionally, specific design choices can lock customers into a single provider’s agent, particularly those related to personalization and memory. An agent that has learned your preferences, work patterns, contacts and past decisions is far more useful than a blank one.

But today, that context is controlled by the provider rather than the user. If firms do not allow context to be exported in a usable format, switching requires the user to start from scratch with their new agent. The longer someone uses an agent, the higher the cost of leaving.

Combined, these factors could drive us towards a concentrated market. They also imply that newcomers like OpenAI and Anthropic still have the potential to become dominant, even though they lack the competitive distribution and ecosystem advantages of the established tech giants.

New risks, known solutions

The fundamental question for the agentic AI industry is simple: whose interests are these agents serving? The more concentrated the market becomes, the less likely it will be those of users and the public at large.

Policymakers can address this problem in two ways.

First, they can attempt to steer the market away from monopoly. This means minimizing lock-in and maximizing interoperability and data portability. We know how to do this in technology markets: it requires open standards and protocols. The success of the Model Context Protocol (MCP), which allows agents to connect to third-party services, is a welcome and promising move in that direction. Regulators can build on this by developing and mandating technical standards that force a market to innovate in a pro-competitive way. The UK’s Open Banking Standard has demonstrated this to great success.

At the same time, Big Tech companies with monopoly power in adjacent markets cannot be allowed to abuse that power to gain unfair advantages in agentic AI. The EU acted quickly and judiciously in Meta’s case, forcing it to allow competitors to distribute AI assistants via WhatsApp. It has taken similar positive steps in the context of Google’s integration of AI into the Android mobile operating system.

Similarly, the UK’s Competition and Markets Authority has recognized that Microsoft’s integration of Copilot into its products is problematic and is exploring interventions mandating interoperability, data portability, and fair choice architecture.

These, however, are but three examples of what we find to be a systemic strategy in the tech industry. Competition regulators must address this epidemic of self-preferencing proactively, before it is too late. Policymakers have already made the “wait and see” mistake several times before in tech. Let us not repeat it with agentic AI.

Competition enforcement alone, however, will not resolve the conflicts of interest inherent in agentic AI. Because of the sensitivity of what AI agents do and the breadth of activities they touch on, regulation must impose transparency and accountability on providers. Users, backed by regulation, must be able to ensure that agents are working in their interests. Legal duties similar to those we have already been applying to human agents for centuries must be adapted for the AI era. This includes disclosures of conflicts of interest, binding duties of care, and bans on the use of sensitive personal data for commercial gain, all imposed on the companies that design and market these agents.

Agentic AI’s future will be shaped by policy decisions taken today

The promise of agentic AI is real. It could help people spend less time on routine tasks and more time doing things they love. More importantly, agentic AI could help rebalance power in the digital economy, including helping publishers secure more value for their content, helping consumers understand and negotiate terms of service for products, and helping people to shape their online experiences according to their preferences, including by cancelling subscriptions, curating feeds, and neutralizing dark patterns and other engagement-maximizing practices.

If regulators get this right, the change could be profound. People would have a capable assistant that works for them rather than for a powerful corporation: one that acts across services regardless of who has built it and that is accountable to its user’s interests. Small businesses and independent creators could reach customers and get paid without routing through a handful of gatekeepers. Time freed from digital drudgery could go to work, family, friends, learning, and hobbies. That future is available, but only if policymakers build it deliberately.

Support Tech Policy Press
If you've found our work helpful, consider supporting us.

Authors

George Colville
​George Colville is a Policy Analyst at Open Markets Institute Europe where he works on market concentration and corporate power in the tech sector with a particular focus on the cloud and AI industries. He is a Talos Fellow, has spent time at the Oxford Internet Institute, and previously worked at ...

Topics

Related

Perspective
We Can’t Monitor AI Agents at Scale. Here’s What It Will Take.July 16, 2026