Nobody Asked Gutenberg for His Sources

Tom Cranstoun · The Machine Experience Authority · CogNovaMX Ltd

Picture Mainz, around 1440. Johannes Gutenberg leans on his press, the screw bites, and out comes a page of text faster than any monk could copy it in a week. It is a genuinely astonishing moment. In that moment, nobody - nobody - leans over his shoulder and says: "Lovely, Johannes. Quick question. Where did you get that paragraph? Who checked it? And if it turns out to be wrong, who exactly is getting sued?"

Nobody asked. There was no one to ask, and no vocabulary to ask it in. The press just printed. It had no byline, no masthead, no editor, no fact-check, and no liability. It was the original ungoverned content-generation engine, and we plugged it in and walked away whistling.

What happened next was not calm. The press poured out text at a speed institutions could not absorb: pamphlets, forged indulgences, anonymous broadsides, propaganda by the cartload. It helped light the Reformation. It got people imprisoned and worse over words nobody could prove they had written. Then, slowly, over about two centuries, Europe built the thing it had skipped: the accountability layer. The byline, so you knew who wrote it. The masthead and the printer's mark - "printed by so-and-so, at such a place, in this year" - so you knew who set the type. Libel law, so there was a name to point at. The Statute of Anne in 1710, so authorship meant something you could hold. None of it shipped with the press. All of it got bolted on afterwards by people mopping up.

Gutenberg, in modern terms, was a startup founder who pushed to production and skipped the compliance review. Move fast and print things.

Here we are again

Generative AI is the press, take two. Infinite text, near-zero marginal cost, no native byline, and the very same three medieval questions standing in the doorway with their arms folded. Who wrote it. Who checked it. Who is responsible for the output. We have better fonts and worse excuses.

This time, helpfully, two camps have formed to argue about the answer, and they are so neatly opposed that you could stage them as a pantomime.

In one corner: the European Union, carrying a stack of Acts thick enough to stop a crossbow bolt. The AI Act, the European Accessibility Act, the Digital Services Act. Their position, stripped of the recitals, is roughly: "Marvellous press, Johannes. Now label it, log it, and give us the name of a human being who answers for it." It is earnest, it is heavy, and on a bad day it threatens to bury the printer under enough paperwork that he gives up and goes back to copying by hand.

In the other corner: the free-speech absolutist, who tends to own a rocket company and a social network, and who holds that any label is censorship, that the marketplace of ideas cleans itself, and that the correct response to bad text is simply more text. Let the press run. Let the crowd sort it out. Tags are for the timid.

Here is the inconvenient part for both of them. The absolutist is proposing the exact experiment we already ran, and we have the receipts: two centuries of chaos before the accountability layer caught up. The marketplace of ideas did eventually sort it out, but only after it invented bylines and libel law to sort it with. The EU, swinging the other way, risks drowning a printer in forms to make him prove what the document could simply state about itself.

So what would Gutenberg have used?

Here is the part that should annoy both corners of the stage: Gutenberg did not need the EU or the absolutist to invent provenance for him. The print trade built its own, and it built it without banning a single word.

Reach into the early printer's toolkit and you find a working set:

• The colophon - the line at the end of the book naming the printer, the place, and the year. "Printed by Fust and Schöffer, at Mainz, 1457." Who made it, where, when. That is the cleanest answer to what Gutenberg's MX would be: a provenance block, set in the artefact, travelling with it.
• The printer's device - Fust and Schöffer's twin shields, later Aldus Manutius's dolphin and anchor. A mark that said this shop stands behind this book. An identity stamp you could recognise across a crowded stall.
• The watermark - and this is the one I would press into his hand. It lives in the paper itself, not in the text. Hold the leaf to the light and the mill names itself. It rides on every single sheet, it survives rebinding, and it is bound to the carrier rather than bolted onto the content. That is the entire MX instinct in a fifteenth-century object: the provenance travels in the substrate, so it goes wherever the page goes.
• The signature marks and catchwords - the small letters at the foot of the page that told the binder which gathering came next. Dull, vital, and precisely the "how to navigate this" layer.
• The imprimatur - a named authority printing let it be printed. A human, on the record, saying this was reviewed.

Lay those over a modern page and you have very nearly the whole of MX: a provenance block, an identity mark, carrier-bound metadata, navigation, and a reviewed-by stamp. Gutenberg had the concepts. He just had to cut them in wood and press them in lead.

The thing he never got

Here is the catch. Gutenberg's world invented MX and never invented Reginald.

A colophon asserts. It does not prove. Venetian colophons were forged up and down Europe precisely because a reader holding the book had no way to check one: you trusted the name or you did not, and the forgers knew it. The imprimatur was no better as evidence - it asked you to trust the authority that issued it, not to verify anything yourself. The watermark came closest, because you could at least hold it to the light, but a determined shop could copy a rival's mark too.

So the print era solved provenance the only way it could, with institutions and habit: guilds, the Stationers' register, reputations built over decades, libel law to punish the liars after the fact. It earned trust by vouching. What it never had was a way for the reader to verify, on their own side, without trusting the messenger.

That is the gap five hundred years finally closes. The modern watermark is one you can check rather than merely read: a signed claim, bound to a resolvable identity, that a reader or an agent tests for itself. Gutenberg gets to keep his colophon. We get to make it un-forgeable.

The bit the sixteenth century actually worked out

Both camps are arguing about a question the printing era already answered, and the answer was not censorship; it was the byline.

A masthead does not tell the printer what he may print. A byline does not gag the author. "Printed by, at, in the year of" removes not one word from the page. What it does is let the reader decide who to trust, which is a service to the reader, not a muzzle on the writer. The accountability layer that calmed the print world down was simply a habit of saying who was behind the words.

That is the whole of the MX argument, and it is much less dramatic than either pantomime corner would like. Provenance is not a content filter but a byline for the machine age. It does not tell the press what it may generate; it lets the page say who set the type, when, and on what authority, in a form a machine can actually read and check. The absolutist can keep his marketplace of ideas - a marketplace still likes a price tag on the goods. The EU can keep its paperwork - paperwork is a great deal lighter when the document fills in its own provenance instead of making a human swear to it in triplicate.

Nobody ever got to ask Gutenberg who was responsible for the output. There was no mechanism, so the question hung in the air for five hundred years while we improvised one. We finally have the technology to let the page answer the question itself - and we are spending the moment arguing about whether asking the question is rude.

It is not rude; it is a byline. Print it, Johannes.

Tom Cranstoun is the Machine Experience Authority and founder of the MX community. He consults on MX strategy through CogNovaMX Ltd.