Somebody Has To Check The AI Passport

This week a big tech company handed its AI agents a passport.

That’s not my word. It’s theirs. ServiceNow calls the new product AI Passport. It’s a digital credential that rides with an AI agent while it works. It says what the agent is allowed to touch and what it is allowed to do.

Now, that sounds like good news. And in one way it is.

Read it slow, the way you’d read the fine print on a loan. The good news is the product. The bad news is what the product admits.

Lets start with what an agent is.

A chatbot answers you. An agent acts for you. It logs in. It moves money. It changes records. It talks to other systems and to other agents. It does all of this without a person watching each step.

The first thing the story admits. These agents are already running.

The article cites a finding that more than half of organizations run agents that can change their systems. And those agents have no identity controls on them. No name tag. No record of what they were cleared to do.

These aren’t toys in a lab. These are working agents inside the companies that hold your bank records, your medical files and your power bill.

The second thing the story admits. The fences leak.

Once an agent is live, the article says, it can quietly go past the permissions it was given. Quietly?

An agent cleared to read your file can end up writing to it. An agent cleared for one system can wander into the next one. Nobody set out to break a rule. The rule just didn’t hold once the machine got moving.

The article points to a pile of bugs in the very connections agents use to reach their tools. Hundreds of them. That’s not one weak board in the fence. That’s a fence with gaps all down the line.

So along comes the passport. It checks the agent at every step, not just at the front door. That is the right idea. Checking at the door and then looking away was never going to work.

The third thing the story admits. And this is the one that should bother you most.

The fix is for sale.

It is a product. The article says so plainly in its own words. It is “a vendor product rather than an open standard.” A company buys it, or it doesn’t. Nobody makes it.

So the companies that pay get the passport. The companies that don’t pay keep running agents with no passport at all. And you have no way of knowing which kind is handling your file.

It was written for the company. Every step it lists is about the company. List your agents. Check your logs. Make sure you can prove things to an auditor. Update your legal plan.

All of that is fine. But look for yourself in that list. You aren’t there.

You are the one whose records the agent touches. You are the one who pays if it goes wrong. And you don’t appear anywhere in the story.

Then there is the record itself.

The passport makes a log of what each agent did. That’s good. A record is how you find the truth after something goes wrong.

But that log belongs to the company. It sits in their system. They decide who sees it. If an agent goes past its fence and into your account, you won’t see that log. You’ll get a letter, maybe. Months later, maybe.

A record nobody outside can read is not accountability. It is a diary.

Here is the last piece. Is any of this required?

Not yet. The article says regulators “may begin” to expect this kind of proof. May begin. That means today, they don’t.

The machines are already loose. The fences already leak. And the fix is optional, private and for sale.

The person in the middle of all this is you. And you are the one nobody asked.

So what can a regular person do?

First, ask. When a company uses AI on your account, ask what it is allowed to do. Ask who checks it. Ask what record it leaves and who can read that record. You have every right to ask. Most people never do, and companies count on that.

Second, keep your own record. Write down what you were told and when. If something goes wrong later, your notes are the only record you control.

Third, stay in the driver’s seat with the AI you use yourself. You set the rules. You decide what it touches. You read what it did before you trust it.

That last one is where the Baseline comes in.

The Faust Baseline is a set of rules for how an AI should conduct itself. One of those rules is plain as day. Before an AI takes an action, it says what record it is leaving behind and who can read it. If it is leaving none, it says that before it acts.

That is the question the passport story never asks on your behalf. The Baseline asks it out loud, on the page, where you can see it.

The Baseline is not a lock. It does not force a machine to obey. It is chosen conduct, written down, so a person can hold the machine to it and see when it slips.

That’s the point. A level playing field doesn’t come from a product you can’t buy and logs you can’t read. It comes from rules you can see and from people willing to ask.

The passport checks the agent.

Somebody has to check the passport.

That somebody has to be us.

Load this file before each AI chat session.

This file is a PDF it will open as text, download it to your computer first.

the-faust-baseline-public-release-1Download

This post was drafted with AI governed assistance and reviewed and directed by Michael S. Faust Sr. before publication.

” Attic Thoughts”-library – Intelligent People Assume Nothing