Meta Is Charging for AI Agents. Nobody Is Watching Them.

Meta launched its Meta Business Agent globally today. It answers customer questions, recommends products, books appointments, qualifies sales leads, and reroutes queries to a human when needed. It runs inside WhatsApp, Instagram DMs, and Messenger. Large businesses pay by the token. Smaller businesses pay through a subscription tier. Zuckerberg is finally diversifying beyond ads.

The monetization story is clean. The accountability story is not.

Here is the problem. When a Meta Business Agent tells a customer the wrong return policy, promises a discount it was never authorized to offer, or mishandles a complaint in a way that creates legal exposure, who owns that? Meta built the model. A third-party platform may have integrated it. The business deployed it. The customer interacted with it. Four parties. Zero clear owner.

This is not a hypothetical. In 2024, Air Canada was held liable after its customer service chatbot gave a passenger incorrect information about bereavement fares. The airline argued the chatbot was a separate entity. The court disagreed. Around the same time, a car dealership deployed an AI chatbot that customers manipulated into agreeing to sell a new Chevy Tahoe for one dollar. In both cases, the failure was not technical. It was an accountability failure. Nobody had defined who was responsible for what the agent could promise, and what would happen when it got something wrong.

Now multiply that problem by the scale of WhatsApp. Two billion users. Hundreds of millions of businesses. Agents that can learn your brand tone, analyze your product catalog, handle customer objections in real time, and eventually negotiate supplier contracts and manage calendars. Goldman Sachs estimates that token use by AI agents will multiply 24 times by 2030. The agents are not coming. They are here. And the accountability infrastructure is not.

The Coalition for Secure AI published a framework this week that maps the problem precisely. They call it the AI Shared Responsibility Framework. Five layers: the business deploying the agent, the data it accesses, the application developer who built it, the platform hosting it, and the model provider who trained it. Every layer has an owner. The question shifts from "whose fault is this?" to "which layer's controls failed, and who owns the fix?"

That is the right framework. The problem is that almost no enterprise deploying AI agents today has mapped their stack against it. They are running agents at levels of autonomy that most governance frameworks do not even recognize as a category.

Meta is not the villain here. They built a product that businesses want and are willing to pay for. The capability is real. The commercial logic is sound. But capability and accountability are not the same thing, and the gap between them is where the liability accumulates.

Every CEO and business leader deploying an AI agent right now needs to answer three questions before they go live. First, what is this agent authorized to promise on your behalf, and who verified those boundaries are actually enforced? Second, when the agent gets something wrong, which it will, what is your escalation path and who owns the remediation? Third, if a customer sues over something the agent said or did, which party in your vendor chain is contractually responsible?

If you cannot answer those questions, you are not deploying an AI agent. You are deploying a liability.

The monetization of AI agents is inevitable and appropriate. Zuckerberg is right to charge for this. The market is right to pay for it. But the companies writing those checks need to understand that they are not just buying a tool. They are taking on accountability for everything that tool does in their name.

The question is not whether to use AI agents. The question is whether you have built the governance infrastructure to use them responsibly. Most companies have not. The courts will not wait for them to catch up.