Monopoly Round-Up: Why the 'AI God' Narrative is Actually a Corporate Power Grab

Source: Matt Stoller; BIG by Matt Stoller (Substack, paid post) · 2026-02-16

If you push the juvenile theological arguments to the side, there are very interesting political questions around machine learning and AI. The fundamental one is whether the companies who build AI tools should make money by helping their customers or manipulating them.

AI is good at certain things, like legal document review, where human beings are modeling and then verifying the output. It is terrible at producing reliable products. And that’s not going to get better, AI algorithms are big guessing machines, they do not “know” anything and cannot understand truth. That’s why attempts to replace, as opposed to supplement, white collar workers doesn’t strike me as likely.

Still, the ability to record and reuse the context and data across our entire society may be magical. If a trusted AI agent had access to your bank account, email, social media, location data, calendaring, home devices, purchases, medical history, search history, and workflow, maybe also a constant video stream of what you see, then weird stuff becomes possible. Apps and user interfaces disappear, and you could set your “agent” or set of “agents” to manage your life or the life of your family in a variety of customizable ways. To give you a small example of how I would use it, I might have an agent watch a bunch of different YouTube channels, and pinging me when someone discusses a monopoly.

Of course, there are many ways this kind of technology can and will be used and misused. Right now, in Silicon Valley, techies are obsessed with an open source tool called OpenClaw, which is the first draft of this personal agent technology. The product is massively insecure and most use cases seem kind of dumb to me, but it’s a notable development. OpenAI hired the creator of OpenClaw, and you can be sure that every big AI company is learning from its deployment.

Taking a step back, every AI chatbot is in a sense a kind of agent, perhaps less sophisticated or capable than OpenClaw, but using the data and context or our chats to offer increasingly personalized service. Creating agents that are accessing our information and using it is a highly political endeavor. Recently, OpenAI made a decision to finance its product with advertising, drawing mockery in Super Bowl commercials. But the joke is deadly serious; if an agent is advertising to us, is it really operating on our behalf? Or is it a conflict of interest?

A few weeks ago, I wrote a piece about how Google is seeking to become the central planner of the economy. It’s always been corrupt to have a search engine funded by advertising, from the very beginning. But after Judge Amit Mehta ruled badly on the search antitrust suit, Google is trying to build out an entire “agentic economy” based on that same strategy. It rolled up personal data, distribution of chatbots, and built partnerships with Apple, as well as virtually every retailer and payment network, in an attempt set pricing across the economy.

When you login to Gmail, Google is now summarizing your email. When you search, Google is summarizing the web. These are prototypes of agents, and they will improve. But are they operating on your behalf? On behalf of the advertisers who pay Google? Or on Google’s behalf? AI is leading to significant changes in search, email, social networking, app stores, advertising, and so forth, Google will have the data and context to create a chokepoint that is even more dominant than the one it has today.

That is, unless we act. One obvious solution is to make sure that all agents have a clear fiduciary obligation to the person who deploys them. To force that, we would have to ban all advertising and all payments to finance agents except fees from the ultimate client. This policy would kill many of the bad incentives that Google has to manipulate pricing and the flow of commerce across the economy, and turn the company into a public utility. Weaker but still useful measures might include banning surveillance pricing or dynamic pricing, or other forms of price discrimination.

There are similar elements when it comes to Anthropic, which is building out market power by focusing on working within regulated spaces. And a host of Chinese AI models are now trying to capture the agent space, since their AI models are generally more efficient than America’s bloated approaches because Chinese firms don’t make money selling cloud computing services. Regardless of who is providing the agent, a regulatory model forcing agents to be on the side of the citizen is very different than one which isn’t. And that’s especially true if Google becomes the full provider of all commercial context and data, or if the Chinese come to dominate in key spheres.

A world where you are manipulated by the company providing you with a window into the world is very different than one where you are just paying for honest services. And that’s what the cultish weirdos are bullying us to avoid thinking about.