AI agents have a big problem: Nobody wants to let them in

Personal AI agents are being marketed as tools that can handle everyday online tasks - banking, shopping, restaurant reservations, and travel bookings - but many businesses are restricting or blocking them. United Airlines says it is limiting unauthorized agents’ access to its platforms and travel intermediaries, citing risks that shared account credentials could expose sensitive information or lead to mistaken purchases, account changes, or use of loyalty miles. It also argues that flight check-in involves a required human acknowledgment that baggage contains no prohibited items, a step an automated assistant may not be able to fulfill appropriately.

The pushback extends across financial services and online retail. Bank of America has warned customers not to give login details to AI tools; JPMorgan Chase, Charles Schwab, and Citigroup have prohibited third-party agents from accessing accounts. Amazon has blocked Meta’s Muse agent, while Yelp, eBay, and The RealReal have also restricted outside assistants. Companies’ motives include protecting customers’ money, preventing bots from disrupting competition for products, and steering users toward in-house AI. The central argument is that agents with credentials and purchasing power are arriving faster than businesses can safely accommodate them. As platforms erect barriers, the promised convenience of agents that autonomously manage bookings, purchases, and finances looks increasingly difficult to deliver.