California did something last week that nobody else has done.
On September 9, the governor signed two AI bills into law. Together they build the first system in this country for having somebody outside a company check that company’s AI.
One law sets up a commission to write safety standards. The other sets up a state registry of AI auditors.
That is the news. Here is the part under the news.
The registry does not open until January 1, 2029. That is over three years away.
And starting that day, nobody can sell you an AI audit in California unless the state has registered them.
But here is the thing. Neither law requires anybody to get audited.
So the state has created a license to do a job that nobody is required to hire out yet.
The inspectors got hired first.
Now I want to show you something from the record, because this is not my opinion.
Back in July of 2025, the California Chamber of Commerce stood up in a committee hearing and opposed the auditor bill.
She said the AI auditing world is virtually nonexistent. No agreed standards for how an audit is supposed to run. No framework for governing how the auditors themselves behave.
Her ask was simple. Let the field grow up first. Let it build its own standards. Then license it.
The state went ahead anyway.
And then they added one amendment
The auditors have to hand over documentation supporting any claim they make about whether their own methods actually work.
The state is about to license a profession, and the one thing it made the profession prove is that its tools do something.
That is the whole question. That has always been the whole question.
I have been writing about this for a year and a half. Not about California. About the gap underneath it.
Anybody can write a set of AI rules. I did. Plenty of people did.
The hard part is not the writing. The hard part is showing the rules changed anything.
Most of what gets called AI governance is a page of good intentions. It sounds right. It reads well. Nothing in it can fail.
A rule that cannot fail is not a rule. It is a wall sign.
So when California says an auditor has to document that its protocols work, California just asked the only question that matters. And it asked it before the standards exist to answer it.
I worked for a promenet cement co. in the cement processing and minig indusry and nothing was allowed and still not allowed without OSHA & MSHA rules adheared to. “Mine Safety Rules” state and federal regulations.
I am not going to tell you this is bad news. It is not.
For years the only person who could check whether an AI system was safe was the company that built it. Grading your own paper. That is finally ending, on paper, in one state.
But watch the order of things, because the order tells you what this really is.
First comes the badge. Then comes the roster. Then, somewhere down the road, comes the standard the badge is supposed to mean something against.
By 2029, the companies this was written for will have had three more years to keep certifying themselves. And the models they are certifying today will be antiques by then.
There is one more thing you deserve to know.
The two biggest AI companies in the country backed these bills. One of them is Anthropic, which makes the AI I use to help draft what you are reading. The other is OpenAI.
When the companies being regulated endorse the regulation, that is not automatically a scandal. Sometimes people do the right thing.
But it is worth noticing. And it is worth saying out loud instead of leaving it in a footnote.
So here is where I land.
California built an office for a job nobody has defined yet. That is not a failure. That is how most real standards start — somebody builds the room before anybody agrees what gets done in it.
The question is who fills that room, and what they bring with them.
Because in three years, a whole lot of people are going to show up in Sacramento holding a folder that says they know how to audit an AI.
And the state already said what it wants to see inside that folder.
Proof the thing works.
” Attic Thoughts”-library – Intelligent People Assume Nothing
Contact: micvicfaust@gmail.com
This post was drafted with AI governed assistance and reviewed and directed by Michael S. Faust Sr. before publication.
© 2026 The Faust Baseline LLC | All Rights Reserved
