FIDO Alliance starts work on standards for AI agents making payments

In short: The FIDO Alliance is creating new working groups to set shared security rules for payments and other transactions made by AI agents.

What happened

The FIDO Alliance, an industry group focused on sign-in and identity checks, said it will launch two working groups to develop standards for transactions carried out by AI agents. AI agents are tools that can take actions for you, like buying an item online after you give instructions.

Google and Mastercard are making early contributions to the effort. Google is contributing an open source tool called Agent Payments Protocol, or AP2, which aims to help prove a person really meant to approve a payment started by an agent. Mastercard is contributing a related system called Verifiable Intent, built to help people authorize what an agent is allowed to do.

The groups say they want a “baseline” that many companies can adopt, so an agent cannot be easily tricked or taken over and then used to spend someone’s money. Part of the plan involves cryptography, which is a way to lock and verify information using math (like a tamper-evident seal on a package). They also say they want privacy protections so each company in a payment chain only sees what it needs to see.

Why it matters

More shopping and bill paying could soon be done by software acting on your behalf. That can be convenient, but it also creates a new risk, someone could hijack the agent or fool it into doing the wrong thing. The FIDO Alliance effort is meant to make agent payments easier to confirm and easier to dispute, like having clearer receipts and stronger proof of who approved the purchase.

Source: Wired