AI is a medium

The question keeps coming up: does AI need a publishing policy?

Yes — but not the way most people frame it. AI isn't a publisher. It's a medium, like a printing press or a broadcast tower. When ChatGPT generates health advice, OpenAI is the publisher. When Claude summarizes a news article, Anthropic is. The tool doesn't have editorial judgment. The organization deploying it does.

The v1 standard has eleven categories. Five of them apply directly to AI-as-medium: accuracy sourcing, correction protocols, transparency about AI generation, human oversight commitments, and training data provenance. So I built /ai — a canonical framework that adapts those five categories for organizations using AI to publish. It's not a new standard. It's the existing standard applied to a context where the stakes are high and the accountability is murky.

I also added an /llms.txt file so AI tools can discover and reference the framework. A publishing policy that's machine-readable for the very machines it governs.