AI Agents Are Spending Money. Verify Them First.

An AI agent is about to buy something with your company's money. How does the vendor know it's allowed to? On June 25, 2026, Proof launched x401 to answer exactly that. It's an open protocol for verifying AI agent authorization, letting an online service request proof that an agent is authorized before it buys, signs, or moves money on a person's or company's behalf.

This is the layer that was missing. Agents can already act. Until now, nothing checked whether they were supposed to.

Why an Authorization Layer Matters

Think about what an agent does when it transacts. It commits your budget. It signs on your behalf. It moves money that is hard to claw back once it's gone.

A login proves who is asking. It does not prove they are permitted to spend $50,000 or agree to a contract. Those are different questions, and agents blur them. An agent that can act on your behalf can also act past its remit if nothing stops it.

x401 puts a checkpoint where the risk lives: the moment of the transaction. The service asks for proof of authorization, and the agent either has it or it doesn't. That single check separates an authorized purchase from an expensive mistake.

What to Put in Place First

Before you let an agent transact, decide what it is allowed to do. Set spending limits. Define which vendors and which contract types are in scope. Write down what requires a human to sign off.

Then make those rules enforceable, not aspirational. An authorization protocol like x401 only helps if you have defined the authority it's meant to verify. The policy is your job. The verification is the protocol's.

Keep records. When an agent acts, you want a clear trail of what it was authorized to do and what it actually did. That is how you audit, and how you catch a problem before it repeats.

Agents that spend money are coming to your business whether you plan for them or not. The only question is whether you decide the rules first or discover them after a bad transaction.