The report by Sally Hubbard, a senior fellow at the Open Markets Institute, argues that policymakers need to help users understand when an AI agent is working on their behalf or when an AI agent is working on their company’s behalf.
As millions of Americans begin to rely on AI agents to answer questions, gather information, shop, and go about their daily lives, a new report from the Open Markets Institute raises what could be the internet’s next defining question.
The report is Who do AI agents work for? Powering and controlling the next InternetA paper by Sally Hubbard, a senior fellow at the Open Markets Institute, argues that as the agent web transforms the economy, policymakers must face a fundamental choice: Should AI agents primarily represent the interests of their users, or should they be allowed to serve the interests of the companies deploying them?
This research is separate from the issue of AI coordination (i.e., how AI behaves according to human intentions). It focuses on the legal principle of agency. An agent is defined as a person who “acts for or represents another person.” Human agents, such as real estate agents and lawyers, have a duty to serve their customers.
“Most people think that AI agents are also working for them.” Mr. Hubbard said. “But how can an AI agent actually serve a user if it is controlled by a company whose profits depend on monitoring, addicting, manipulating users with targeted content, or collecting fees?”
The report claims that some already dominant technology companies, including Google, Microsoft, Meta, and Amazon, are incorporating AI agents into their existing proprietary products and services. However, these companies have always worked primarily for shareholders rather than users.
The potential for abuse is high.
-
When an AI agent helps consumers buy airline tickets, how do we know that the agent is finding the best deal, rather than charging extra and discounting it?As supervised pricing becomes more prevalent, AI agents may know more about what people are willing to pay than the sellers themselves.
-
When an AI agent answers questions about politics, public affairs, and breaking news, how can users know that the information is not being filtered in a way that promotes hidden commercial or political interests?
-
When an AI agent learns about a person’s finances, relationships, fears, health concerns, habits, and vulnerabilities, who is ultimately in control of that information, and for what purpose?
The report alleges that Big Tech companies have interposed themselves as Internet intermediaries between readers and content creators, and between buyers and sellers. As a result, the fruits of the labor of artists, merchants, publishers, businesses, and workers were extracted.
AI agents are the next internet intermediaries, potentially amplifying the harms of the digital age by fueling exploitative business models. But it doesn’t have to be this way.
“AI agents created by new innovators could help arm artists, distributors, publishers, companies, and workers against harmful business models.” Mr. Hubbard said. “Policymakers can actively encourage the development of a decentralized web of agents that works for the public. The choice is ours.”
But if policymakers do nothing, the choice will be ours. The report offers several recommendations for policymakers.
“Agent Web will transform the Internet as we know it.” Mr. Hubbard said. “Policymakers must act now to ensure that AI agents that serve people are not crushed by dominant companies with exploitative business models.”
###
