The Signal · Column 029

Europe just made provenance a hiring problem

Andrei van Noordt · 17 August 2026 · 1 min read

Europe has decided that the most useful thing a piece of content can carry is an honest answer to a simple question. Europe's new AI rules push toward disclosure, so that if a system generated what you are looking at, someone is expected to say so. Most of the coverage reads this as a consumer story. A label on a video, a watermark on an image, a small grey line under a headline. That is the smallest part of it.

The part most people miss is where the work actually lands. A disclosure is a claim a company makes about its own systems, and you cannot make that claim from the marketing department. Someone has to know which model produced which output, on what data, reviewed by whom, and be able to show it months later without a week of archaeology. Plenty of companies shipped AI into their products at speed over the last two years. Far fewer kept the paper trail that a rule like this quietly assumes already exists.

Regulation did not create a compliance job. It revealed which companies could still remember what their own systems did.

So the market reprices again, and in a direction that surprises people. The scarce hire stops being the person who can wire a model into a product, because that skill has spread quickly and cheaply. It becomes the person who can account for one. Data lineage, evaluation, versioning, the unglamorous discipline of writing down what a system does before anyone asks. Those skills sat at the boring edge of the interesting work and were paid accordingly. Rules have a way of moving the boring edge into the middle of the room.

For hiring leaders, the practical read is simple. Before you add another engineer who can build with AI, secure one who can explain, in writing and on demand, where any given output came from. That person is inexpensive while the requirement is still theoretical, and expensive in the week it stops being theoretical.