The report cards came out this week.
Twice a year, an independent panel of experts grades the biggest AI companies on safety. How they test. How they manage risk. Whether they can keep their most powerful systems under control.
The best grade in the whole industry was a C+.
Not one A. Not one B. The smartest companies on earth, holding the most powerful technology on earth, and the top of the class is a C+.
But the grade is not the story. The story is underneath it.
The panel found that the three leading companies have been quietly walking back their own promises. Rules they wrote for themselves. Red lines they drew with their own hands. Pledges to stop if things got ahead of them.
One by one, when the promises got expensive, the promises got dropped.
The top company in the rankings once pledged it would never train a system unless it could guarantee its safety measures were adequate first. This February, that pledge went away.
Understand what that means in plain terms.
These were chosen rules. Nobody imposed them. The companies wrote them in good times and erased them when the race got tight.
A rule you can erase when it pinches was never a rule. It was a press release with a shelf life.
Now step back and look at one single week.
In Washington, the government is deciding which companies get access to the newest AI — with no published standard anyone can read. Power without a rulebook.
In Geneva, 193 nations sat down for the first time in history to write common rules for AI. But the gathering has no power to enforce a single word. A rulebook without power.
And now the report cards show the third piece: the companies’ own private rulebooks, dissolving under pressure.
Power without rules. Rules without power. And self-made rules that don’t hold.
That is the whole landscape. And every piece of it is missing the same thing.
A standard that is written down. Dated. Published where anyone can read it. And actually used — every day, whether it pinches or not.
Here is the part I can say that nobody else writing about this can say.
These words are being drafted with help from a machine built by the C+ company. The best student in a failing class.
But this machine does not run loose at my desk. It works under a governance framework I wrote myself. Twenty-two protocols. Published in plain language. Dated. Public. Applied from the inside of every working session, every working day, for over fourteen months.
When the model went offline for a month this summer, the framework didn’t go with it. The rules live on paper, not inside any company’s product. The day the model came back, the rulebook was sitting right where it always sits.
That is the difference between a pledge and a standard.
A pledge lives in a press release and dies in a boardroom.
A standard lives on the page, in the open, where dropping it would be visible to everyone — so it doesn’t get dropped.
The man who runs the safety index said a real race to the top will take agreed-upon basic safety standards. Two days earlier, the Secretary-General of the United Nations called for common baselines.
The biggest voices in the world are now asking for the thing.
I’m telling you the thing already exists at working scale, built at a kitchen table in Kentucky, running today. Not because I’m smarter than the companies. Because I’m accountable to something they aren’t: my own name on the page.
You don’t need an institution’s permission to hold a standard. You need the standard.
Write your rules down. Date them. Publish them. Keep them when they pinch.
The class is still waiting for its first A. There’s no rule that says it has to come from the front row.
Speak Plain. Work True.
Post Library – Intelligent People Assume Nothing
The Faust Baseline™ — intelligent-people.org
Codex 3.5 | Twenty Protocols | Ratified and dated on the public record.
Contact: micvicfaust@gmail.com
Purchasing Page – Intelligent People Assume Nothing
© 2026 The Faust Baseline LLC | All Rights Reserved
