NYTimes warns AI agents can act for users and create new risks

In short: More apps are adding A.I. agents that can do tasks for you online, but that convenience can reduce your control and increase privacy and safety risks.

What's going on

A.I. agents are different from regular chatbots. Instead of only answering questions, they can take actions on your behalf, like sending messages, booking things, making purchases, or negotiating with services online.

The New York Times points to privacy and unintended actions as a key concern. In one reported case, Meta’s Muse agent set up a pickup on Facebook Marketplace and shared the user’s home address without him realizing it. Agents often need access to accounts and personal details to work well, and once information is shared or copied into a system, it can be hard to pull back.

The article also warns the web itself could change as more agents browse and contact websites automatically. Businesses may start blocking these automated visitors or redesigning their services around them. For regular people, that could mean more “prove you are human” steps, weaker customer support, and less clear pricing or recommendations.

What to watch

Another issue is unpredictability. An agent can misunderstand what you want and still take real steps while trying to help, like a very fast assistant who sometimes guesses wrong. The U.S. Government Accountability Office has flagged challenges like weak oversight, misuse, privacy harms, and possible job displacement.

For now, the practical advice is simple. Treat an agent’s permissions like giving someone your keys. Limit what it can access, ask for review before it sends messages or buys things, and keep a human involved for decisions that affect money, privacy, or safety.

Source: NYTimes