Watermarked AI text puts a price on the human signature
OpenAI is preparing to watermark the text ChatGPT produces for users in Europe in order to meet the transparency obligations of the EU AI Act. No clever detection tool, no arms race with students. The model marks its own work, because a regulator asked for it rather than because a customer did. Whatever you make of the mechanism, this is the moment machine authorship stops being something you infer about a document and starts being a fact attached to it.
Most of the discussion I see treats this as an academic integrity story, which misses where it actually lands. It lands inside companies. Once generated text carries its origin, the interesting question in any workflow is no longer whether AI was involved. It is who reviewed it, who corrected it, and who is prepared to put their name on the result. Provenance moves from something you can shrug about to something that shows up on the page.
Once every machine-written sentence is labelled, the scarcest thing in a document is the name of the person willing to sign it.
That is a repricing, and it moves in a predictable direction. Volume production of plausible text was already losing value, and a label on it only accelerates the decline. What gains value is the part that cannot be stamped by a machine: the reviewer who knows which third of a draft is wrong, the domain expert whose judgment is the thing actually being certified, the engineer who can build traceability into a pipeline instead of writing a policy about it. Compliance literacy stops being a legal specialism and becomes a technical hiring requirement for anyone shipping AI into the European market.
For hiring leaders, the practical read is simple. Stop testing candidates on their ability to produce work a model now produces faster and pre-labelled. Hand them machine written output in the interview and ask them to tell you what is wrong with it. The people who can do that fluently are the ones whose signature will be worth something once every other sentence in your organisation arrives already stamped.
Planning a senior AI or data hire and want the market view first? We share it either way.
Send us the brief →