Veritas Silentii Conversations Across Time
Conversations Across Time
Conversation No. 7 · October 2026
Friar Bonaventure and the Contributors to a Written Book
The thirteenth-century answer to a 2026 argument about AI-written text. Whose words, whose ideas, whose creation.
During a week in September 2026, a newspaper ran three AI-writing detectors through a test.
All three identified text that was entirely AI-written. All three also cleared other text that they deemed was entirely human-written. Then the testers mixed the two, and the tools lost credibility. One missed the mark badly by flagging a sample that was one-quarter AI-written as almost 90 percent. And human writing that was polished with Grammarly passed as completely untouched by AI.
A lawyer was quoted in the news story…he thought about the result and suggested that employers need to be careful. He told them that even if they could claim computer programs conducted the analysis, "the business still owns the decision" that results from it.
It's a caution that should concern all of us. Think carefully about what actually happened.
Software reads the essay and its algorithm produces a number. A business manager who cannot tell machine prose from human prose…because almost no one can…acts on the machine's conclusions.
The business makes a decision no one made.
We will come back to that. But first, an older problem lies beneath it…the AI detectors aren't failing at their job. They are doing the wrong job. And machines can't do the job that needs to be done.
The question the detectors cannot ask
Every detector analyzes only the writing. It looks at sentence rhythm, word choice, and how predictable the next word is. Think of it as reading the statistical fingerprint of the hand that typed it.
But the hand that typed it may not belong to the person who had the idea. Like a page produced by a typewriter, no text carries that information. It never has.
And whose idea it was is the missing information that we need if we want to fully analyze all the issues surrounding AI-created text.
Stop reading for a moment and think about how ownership works everywhere else in our lives. If I have an idea for a product and hire engineers to build it, the product is mine, and everyone involved knows it. If I run a small creative firm, and I work with a client to develop a design brief that my staff develops and executes under my direction, the work carries my name. Nobody ever would call it fraud, because nobody confused the idea with the execution.
The problem today with AI writing is that we have collapsed those two things into one. We seem to be saying that whoever typed the words settles the argument about who thought the thoughts.
It doesn't, it never did, and a very long time ago we knew that, not only in detail, but with agreed-upon rules.
However, we seem to have forgotten, and I think it's because for the past two centuries the person with the idea and the person holding the pen were nearly always the same person. AI has pulled them apart again, but those old rules are lying there in plain sight, right where we left them.
That's what the article is about. So let's go back by almost a millennium.
Paris, 1250
In the middle of the thirteenth century, a Franciscan friar named Bonaventure sat down to write guidelines for creating the standard textbooks of his day. Of course, in his day, they were almost always theology textbooks. He began, as scholars always do, by carefully defining the thing he was making.
We need to remember that Bonaventure lived in a world where the hand and the mind were routinely different people. Books were copied by scribes, arguments were developed by church leaders, and a single volume might contain the words of six men and the thoughts of several others.
So, the question of who had created the textbook was a critical one, and Bonaventure answered it by classifying four sources that contribute to making a textbook.
- The scriptor writes another's words, adding and changing nothing. He is a copyist.
- The compilator writes others' words, arranging them, but adds none of his own. He is a compiler.
- The commentator writes others' words and adds his own, but only to explain theirs. He is a commentator.
- The auctor writes his own words and adds others' only to confirm them. He is an author.
These four tiers define not only who held the pen but also whose words and whose ideas were on the page, and in what proportion.
Notice something that is especially relevant for today's arguments…the classifications do not contain a tier for "wrote every word himself with no help." That was so rare it didn't need a name. In addition, it has no tier for fraud, because a scribe who copied well was doing honest work. Bonaventure would have defined sins as things like claiming to do something you didn't, such as being an auctor when you were a scriptor.
Seven hundred years before anyone needed it, Bonaventure had written the rulebook for AI-assisted writing. But I suspect that almost nobody outside a medieval studies department has read it. Hopefully, this article can help to change that.
Now let's move ahead about four hundred years.
Antwerp, 1618
The famous painter of the Flemish Baroque movement, Peter Paul Rubens, ran a studio in Antwerp that produced far more paintings than one man could paint…and he never pretended otherwise.
On April 28, 1618, he wrote to Sir Dudley Carleton, the English ambassador at The Hague, proposing to trade paintings from his studio for Carleton's collection of antique statues.
He even made a detailed list of his offerings, where each painting was described by how much of Rubens's personal work was in it. Some paintings were by Rubens's own hand, and others were started by a pupil and finished by him. And still others had been created by pupils following his designs, then retouched by him. Rubens set the price for each of these tiers.
Carleton read the list and, with this information about authorship, he chose as most buyers would if they had the necessary financial resources. He wanted only the ones entirely by Rubens's hand and said so…and the deal was negotiated around that.
Nothing in this negotiation related to who produced the paintings was scandalous, because it was how a great studio worked, and everyone who mattered knew it. The idea, the composition, the standard, the final touch were Rubens's, and that was what his name meant. But the execution was also a factor that was disclosed, tiered, and priced.
Carleton did not need a detector, because the painter told him.
Paris, 1858
Two hundred years later, those same rules were still working, and a French court said so.
For example, most people today will be surprised to learn that Alexandre Dumas produced novels at a rate everyone at the time knew that no one man could sustain. He didn't, and he didn't lie about it. A schoolteacher named Auguste Maquet actually drafted much of The Three Musketeers and The Count of Monte Cristo, working from Dumas's plots, characters, and direction. In a process that might sound very familiar to many of us today as we work with AI-generated text, Dumas rewrote these drafts, added his thoughts, cut others, and then signed the work as his own.
In 1858, when Dumas's books became well known, Maquet challenged this arrangement and sued to have his name on the books.
The court found that he had been paid for his work, and that the work he had been paid for was execution. The idea and the direction were Dumas's and Maquet lost the case. And today Dumas is on every shelf, and almost nobody has heard of Maquet.
It is the nearest legal precedent to the argument we are having now, and it was decided the way most owners of a creative idea would want: the author is the one who created the idea and who directed the execution.
The rule protects the mind, and carries a cautionary tale for the hand.
How the rule got lost
Somewhere in the nineteenth century, the hand and the mind moved into the same body and stayed there. And it's the reason for today's confusion.
It's worth noticing that this is the opposite of what the industrial age did to everything else. Take goods, for example. In the old world it was one mind, one hand, and one body. A craftsman made the whole chair, and the guild existed to keep it that way. Then, the industrial age factories broke that apart with multiple operations run by different people, and nobody actually making a whole product themselves.
Writing ran the other way. Before the industrial age, so few people could read or write that the hand and the mind were routinely owned by different people. So when authors dictated and scribes copied, no one was confused. It was the same for art forms where masters signed what pupils painted. That is why Bonaventure needed four tiers, and Rubens needed a tiered price list.
The printing press was the first machine to upset that arrangement, and it is worth noticing what it did. It did not join the hand to the mind. Instead, it replaced the scribe. The author went on dictating and drafting while the copying passed to a machine, and the compilator and commentator became the editor and the publisher. And no one in 1500 was confused about who the author was.
What the printing press did change was everything around it. Cheap books made mass literacy possible. But there was a downside too. Industrial copying made industrial piracy possible, and the law answered in the Statute of Anne of 1710 by compressing three of Bonaventure's four tiers into one person, the author, who owns the text. Then, somewhere in the nineteenth century, the typewriter finished the job, and the person who had the idea, the author, became the person who typed it.
The Romantic idea of the writer as a single creative source, thinking and typing in one motion, became the only picture anyone had.
During that time, cheap paper, mass literacy, the typewriter, and the word processor made execution so cheap and easy that a writer could do it alone. And for two centuries the earlier classifications of tiers slowly fell out of use. Nobody needed a word for the compilator when everyone typed their own work.
So, the two histories cross during the industrial era…making of things was split between mind and hand, and for the writing of words, they were fused.
The confusion people feel today is because AI is now looping writing back to the older arrangement where the hand and the mind were different people, everyone had words for it, and no one was confused. We have not adjusted yet. We will. And I hope this article will help.
There is a curious place where the old rule has survived. In the field of science, a modern paper may have a dozen authors, and the journals long ago stopped pretending that "author" meant one thing. They now require contribution statements based on standard classifications that were adopted in 2014. They list fourteen roles, and two of them are the whole argument of this essay in miniature.
Conceptualization is one role, and writing the original draft is another. Contributors are another. They are routinely different people, and nobody is confused about who owns the idea.
Bonaventure would agree with this at once.
Three purposes, three rules
I think the reason for the current argument going in circles is that it applies one rule to three different situations. Rubens, Dumas, and a schoolboy with a term paper are not doing the same thing, and in medieval times we would never have mistaken that they were.
- When the purpose is to communicate an idea, ownership follows the idea. This essay is an example. The thinking is mine; a machine helped with research, drafting, and checking, the way a professional staff would help a creative leader with design and production of an advertising program. The honest way to handle this is the way Rubens did. Just define the tier, not because the hand determines ownership, but because the reader is owed the truth about how the thing was made and will trust you more for it.
- When the purpose is to certify that a person can do the work, the execution is the point. A student's essay exists to prove the student can think and write, and a bar exam exists to prove the candidate can reason like a lawyer. In this case, the machine's help is not fraud in the ordinary sense; it is something worse and more specific. It defeats the purpose of the exercise. This is Whitehead's territory, not the copyright lawyer's. An idea the student never used, tested, or combined is useless cargo in his brain, despite how good the essay looks.
- When the purpose is the craft itself, the hand is the product. A poem, a painting, a piece of furniture bought for the maker's touch. Here the buyer is owed the description of the tier above all, because the tier is what they are paying for. Carleton wanted Rubens's hand and said so and got it.
The current idea of AI detectors, and the policies built on them, apply the second rule to everything. And that's the problem. It's why a business essay written the way businesses have always written things, by a person with an idea and help by his staff with the execution, gets treated like a plagiarized term paper.
I want to make one caution clear, because a fair critic will raise it. While your staff may have executed your idea, they consented, were paid, and could be credited. The AI system's executional skill, on the other hand, was distilled from other writers' work with none of those things. So the real issue is in two parts. The first part is a legal issue about who owns the idea. The second part is essentially a grievance about where the ability came from to execute the idea. Those two arguments are each valid, but stronger and less confusing if kept apart and not tangled together.
The decision nobody made
Now let's get back to the lawyer's rule that the business still owns the decision.
He was pointing out a somewhat confusing paradox. The judgment the business is enforcing is one that humans cannot reliably make or validate, so in effect, they are relying on machines they don't fully understand to make judgments about machine writing that they don't fully understand. Researchers have tested it repeatedly since 2021. Ordinary readers sort machine writing from human writing at close to a coin flip, and the cues they lean on…personal pronouns, contractions, mentions of family…are exactly the cues the models have learned to produce.
So, when the business faces a judgment no person on its staff can perform, it hands the judgment to a machine that returns a probability. A person who cannot check it treats it as a verdict. And the person on the other end may be punished or validated by a machine.
So, as the lawyer in the news article suggests, the business still owns the decision no one made. I have a name for that condition: Agentic Inertia, where everything runs perfectly well without you, and no one, least of all you, chose any of it.
However, I think there is a deeper version of the problem, and it is the one that should worry the detector companies.
A detector is trained on a picture of human writing drawn from the past. But the newspaper's own test showed Grammarly-edited prose passing as fully human, which means the picture is already moving. Every essay that passes through a machine on its way to a page, and is then filed as human, shifts the baseline.
Think about it this way. The machine that writes must keep learning from people who can still write without it. The machine that catches writers who pass off machine prose as their own learns from the same people. Both are living on one supply…writers who can still write…and both seem to be using it up. And no one really knows yet what happens then.
The Bonaventure Echo
The way out is older than the problem, and it does not require a detector.
Just declare your tier.
Rubens did it on a price list. Scientists still do it in a contribution statement. A friar did it in 1250 before he had written a word of his own book. It costs one sentence, and it moves the question from the machine, which can only guess at the hand, to the author, who alone knows the idea.
For anyone who writes with help, three habits follow.
- Say what the machine did. Drafted, checked, polished, nothing. One line at the end. Readers forgive almost anything they were told and almost nothing they discovered.
- Know which purpose you are serving. If the point is the idea, the tier is just a courtesy. But if the point is to prove you can do the writing or the thinking, the tier is the whole test.
- And for anyone who judges the writing of others: stop asking the machine who typed it. Ask the writer to defend the idea they claim to own. A person who cannot does not own it, despite how clean the detector score.
Bonaventure never met one of our machines. But he would have had no trouble questioning their purpose.
Whose words? Whose idea?
What Inspired This Conversation
This Conversation draws on Bonaventure's prologue to his Commentary on the Sentences (c. 1250–1252), where the four ways of making a book are set out, as discussed in A. J. Minnis, Medieval Theory of Authorship (1984); on Rubens's letter to Sir Dudley Carleton of April 28, 1618, in Ruth Saunders Magurn, The Letters of Peter Paul Rubens (1955); on the Maquet v. Dumas litigation of 1858; on the CRediT contributor roles taxonomy (2014); on studies of human detection of machine text, including Clark et al., "All That's 'Human' Is Not Gold" (ACL, 2021) and Jakesch, Hancock and Naaman, "Human heuristics for AI-generated language are flawed" (PNAS, 2023); and on the Epoch Times test of three AI-writing detectors published September 30, 2026, with the quoted remarks of Michael McCready of McCready Law.
A note on how this article was created, in the spirit of the essay: the ideas and argument are mine. An AI assistant helped draft, check sources, and edit. Bonaventure would call me an Auctor with a fast Scriptor.