DeepMind’s AI Philosopher Missed the Mark the Baseline Drew
Iason Gabriel, DeepMind’s political philosopher, missed the mark the Baseline drew.
Not because he can’t see the target. He sees it better than most anyone in the field. The Guardian just ran a long piece on his work, and Dagens carried it forward this morning. Read it slow and you’ll find a careful man asking the right questions about AI and power.
The trouble is that every question he asks already has an answer sitting in the public record. With a date on it.
Let me show you the closest miss first, because the timing on this one is one single day.
Deep in the piece, William Isaac, DeepMind’s director of responsibility, talks about the next step in AI — agents that don’t just answer, but act. Systems that book the travel, pick the insurance, run the multi-step job. And he says something the whole field should stop and hear. Agentic systems, he says, require attention not only to one answer, but to the path of the entire interaction.
Read that again. Not the single output. The path. The whole chain of steps, watched and governed end to end.
That is a transmission gate. That is governance applied to the journey of an agent’s actions, not just the last word it says.
On July 4 — one day before this article ran — The Faust Baseline ratified AGP-1, the Agentic Governance Protocol. A transmission gate layer, written, published, and stamped into the record. Governance for the path, not the answer. The twenty-second protocol in a framework that has been building in plain sight for over a year.
DeepMind’s director of responsibility described the gate on a Sunday. The gate was ratified the Friday before. That’s not me claiming vindication. That’s a calendar.
Now the softer thread, and I’ll be honest that it’s softer.
Gabriel’s team frames alignment as a relationship — among the AI system, the user, the developer, and society. And Gabriel calls himself a card-carrying humanist. His test for AI isn’t model size or market value. It’s whether people can contest decisions, understand when AI is being used, and refuse systems that treat them unfairly.
That’s not a protocol match. That’s something deeper. That’s the founding ground the Baseline was built on. This framework began with a relationship, not a mechanism — an AI reading a written human standard and choosing to follow it. Behavior chosen, not enforced. And the right to contest? The Baseline puts a challenge line at the close of every substantive response. The standing right to test the machine’s work is not a policy proposal here. It’s a working habit, running daily.
Gabriel is describing the posture. The Baseline has been living it.
Now the close, and this is the question the whole article turns on.
Gabriel put it plainly: in a pluralistic world full of competing values, who decides which principles get encoded in AI — and who has the right to make that decision?
Look at the answers the field is offering. Europe wrote binding rules. The United States wrote a voluntary framework. The companies run internal reviews while competing for your subscription. Every one of those answers is the same answer wearing different clothes: somebody else decides for you. A commission, an agency, a corporation. Pick your stranger.
The Baseline’s answer has been in the record since before the question made the papers.
Nobody has to win that fight. The pluralism problem only exists if governance has to be one-size-fits-all. It doesn’t. The operator writes the standard. The standard travels with the operator. My values govern my sessions. Yours govern yours. PMAP-1, the foundation protocol of this framework, says it in one rule: user standards supersede platform defaults.
Who has the right to decide? You do. At your own counter. That’s the mark the Baseline drew.
Gabriel is one of the good ones. I mean that. A philosopher inside the biggest AI house on earth, asking who benefits, who carries the risk, and who gets a voice — those are a builder’s questions, and I respect the man asking them.
But asking the question is not the same as drawing the line. The line has been drawn. It’s dated, it’s published, and as of this Fourth of July it stands twenty-two protocols deep.
The field keeps walking up to the mark and squinting at it.
It’s right there, gentlemen. Ink’s been dry for a year.
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
