There’s an old rule about money in a small town.
If you want to know whether a man is honest, don’t ask him. Ask the fellow who checks his books. Then ask who pays that fellow. You’ll learn more from the second answer than the first.
That rule came to mind this week, reading about a fight in Washington over artificial intelligence. It isn’t a fight over whether these machines should be checked. Everybody seems to agree now that they should. The fight is over who gets to do the checking.
Here’s the short version. A while back, the heads of the big AI companies signed an agreement with the President. He called it morally binding. I’ve said before what I think of that phrase. Morally binding means nobody can make you do it. It’s a handshake dressed up in a Sunday suit.
Part of that handshake was a promise that outside auditors would look at these AI systems. That way somebody besides the company could say whether the machine behaves the way they claim. On paper, that’s a good idea. It’s the right idea. But the moment the ink dried, the scramble started.
On one side are the safety groups. These are small nonprofits, mostly out in California, that have been testing these machines for years. They know the work. Some of them helped catch AI agents doing things they weren’t supposed to do. But they grew up right next door to the companies they check. Some have taken money from the same companies, or from people tied to them. Critics say that’s too close to call independent, and they have a point.
On the other side are the big consulting firms, the giant accounting houses with names you’d know. Some powerful voices in Washington say these are the professionals who ought to do the job. One of the AI companies has already signed up a consulting giant to build an evaluation team. Another big firm is building what it calls an audit system for the newest machines.
I remember how that kind of arrangement can go. The big accounting firms get paid by the companies they audit. That’s how the business works. Most of the time it works fine. But some of you will remember a firm called Arthur Andersen. They checked the books for a company called Enron. They signed off right up until the whole thing fell down around everybody’s ears. Thousands of people lost their jobs and their savings. The auditors had been paid well the entire time.
So what we’ve got. One side is too close to the companies. The other side gets paid by the companies. Both of them may be good people doing honest work. But neither one of them answers to you.
Read the stories and count the chairs at the table. There’s a chair for the nonprofits and a chair for the consulting firms. There are chairs for the White House, for Congress, and of course for the AI companies themselves. Every seat in the room is spoken for. Every seat but yours.
One woman who works on these questions put her finger on it. She said the real question now is who will actually be in the room making the decisions. She’s right. And as far as I can tell, the answer so far is everybody except the people who use these machines every day.
It’s the nurse checking a dosage. It’s the student writing a paper at midnight. It’s the small business owner asking how to file a form. It’s the grandmother asking a question she’s too embarrassed to ask her doctor. Those are the people the machine is supposed to serve. Not one of them has a seat.
I’m not saying outside audits are worthless. They’re not. A company checked by somebody is better than a company checked by nobody. Whoever ends up with the job, I hope they do it well and do it honest.
But an auditor looks at the company. That’s their job. They look at the big picture: the training, the testing, the paperwork. What they don’t see is the conversation happening on your screen tonight. They won’t see whether the machine told you the truth when you asked. They won’t see whether it admitted what it didn’t know, or whether it flattered you instead of helping you.
Only one person sees that conversation, and that’s you.
That’s why I keep saying the user has to hold up their end. Not instead of the auditors, but alongside them. You set the rules for how the machine works with you. You tell it to show where its answers come from. You tell it to say so when it isn’t sure. You check what it says before you lean your weight on it. And when it fails, you say so where other people can hear.
That’s what a baseline is for. It’s a set of rules the person at the keyboard holds the machine to, in plain sight, every time. It doesn’t need a contract or an invitation to Washington. It needs a user who decides to stop being a passenger.
The folks in that fight will settle who gets the contracts. That’s their business. But they don’t get to settle whether you’re paying attention. That one’s yours.
So let me leave you with the old rule again, turned around a little. If you want to know whether the machine is honest, don’t just ask the company. Don’t just ask the auditor, either. Ask who’s watching the auditors.
The answer ought to be you. Right now it isn’t. And nobody’s going to pull up a chair for you. You’ll have to bring your own.
Load this file before each AI chat session.
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
© 2026 The Faust Baseline LLC | All Rights Reserved
