The New Hunting Ground

The hunter has found new quarry. For two decades the Austrian “plagiarism hunter” Stefan Weber tracked the copied passages in politicians' doctoral theses and the lifted paragraphs in the books of public figures; his trade was the proof of foreign words in one's own text. With artificial intelligence the game has changed, but the hunt has stayed the same. No longer is it the foreign word that is sought, but the machine-made one. This week Weber confronted the former Tagesspiegel editor-in-chief Stephan-Andreas Casdorff: he suspected that the man's opinion pieces had been written by an AI, set a forty-eight-hour deadline, and the journalist admitted to having used AI without marking it as such. He was provisionally barred from writing. A few days earlier the FAZ had deleted a text by Thuringia's minister-president on suspicion of AI use; in it Weber saw confirmation of the “intellectual emptiness” of the political and media elite, a “text culture without a brain”.

We mean to take this seriously, and for that very reason we must take it apart. For what appears here as one charge is in truth three. Deliberately or not, they are tangled into a single knot, and only thus does the accusation gain its force. Separate the threads, and less remains of the moral outrage than it promises — while at the same time a real core remains that we do not wish to talk away.

The First Charge: The Concealed Origin

The first charge is justified, and we concede it without reservation. When a respected journalist publishes under his own name a commentary that a machine has written, leaving the public to believe that here he himself is thinking and formulating, then he deceives. The reader of an opinion piece is not merely buying sentences; he is buying a judgement to which he attaches a name, a biography, a responsibility. Undercut that attribution, and trust is violated — quite independently of whether the text is good or bad. This is not about quality but about honesty regarding origin. Casdorff himself put it this way: he ought to have marked the use. That is the hard, load-bearing core of the whole affair, and whoever disputes it is making things too easy for himself.

The Second Charge: The Unnamed Source

The second charge is finer, and it strikes a spot where one does not expect it — namely the comfortable defence that disclosure alone already cures everything. The legal scholars Mark Lemley and Lisa Ouellette have spelled it out: AI is a “plagiarism machine”, but not in the sense of copyright. It rarely copies foreign phrasing verbatim; yet it draws its ideas from somewhere and is unable to name that origin. Whoever works with it may therefore pass off foreign ideas as his own without knowing it. And now the uncomfortable point: the mere note “AI was involved here” does not solve this problem, for the originators of the underlying ideas still receive no credit by it. We call this uncomfortable because it aims at our own favourite intuition — that transparency settles the matter. It settles the deception about authorship. It does not settle the quiet drawing from unnamed wells. That is a real deficit, and it demands more than a label: it demands the labour of following the tracks of the ideas back to their source.

The Third Charge: The Inference from Origin to Worth

Only here, with the third charge, does what actually concerns us begin — and it is the weakest of the three. It lies in the tone, not in the argument: in “text culture without a brain”, in “intellectual emptiness”, in the English epithet slop, in the whole gesture of contempt. Here it is no longer claimed that someone deceived or withheld a source. Here it is claimed that the result is worthless because a machine was involved in it. That is no longer an integrity argument but an aesthetic prejudice dressed up as integrity. Logically it is a genetic fallacy: the inference from the manner of a thing's making to its worth. A thought does not become false because a pocket calculator took part in its arithmetic, and a sentence does not become stupid because a machine helped to shape it. Tellingly, the more serious observers avoid this third charge. Some recall that today's slop debate is merely the rerun of the hundred-year-old quarrel over ghostwriting — a practice always tolerated, provided the author took responsibility. The academic institutions draw the same line: most major journals permit AI assistance, provided it is disclosed; what is forbidden is not the involvement but only that the machine appears as author and the human evades his responsibility. The third party's contempt for everything machine-assisted is thus not the consensus but a presumption.

The Contaminated Instrument

That the hunter conflates three charges would still be forgivable if he pursued his hunt with a reliable tool. He does not. The AI detectors on which the new evidence rests are expressly not truth machines; they report probabilities, not facts, their providers warn against using them as the sole basis for penalties, and false positives are the rule, not the exception. Even Weber concedes that all these tools are to be treated with caution — and bases his confrontations on them all the same. This is accusation by conjecture, delivered with the authority of measurement.

And the hunt is not neutral. Weber's targets bear names that form a pattern: Robert Habeck two weeks before a federal election, Kamala Harris three weeks before a presidential one (the American plagiarism researcher Jonathan Bailey called those charges “not serious”), now a minister-president — and joint appearances with the FPÖ and the far-right Thuringian AfD politician Björn Höcke. Across the Atlantic, the investor Bill Ackman announces that he will use AI to comb entire academic life's-works retroactively for missing quotation marks: no work in academia, runs his threat, survives the power of the AI search. This is the true nature of the new hunting ground. An existing apparatus of takedown — politically motivated, built on attention — has acquired a new, cheap, unreliable tool with which almost anyone who has ever written something can be hit. Whoever can declare everyone guilty has ceased to ask after guilt.

The Fourth Possibility

There remains the strangest thing — what the whole discourse cannot see, because its concepts have no place for it. The dispute knows exactly two figures: the tool one dutifully declares, and the fraud one conceals. On one side the machine as a typing aid, to be mentioned in a footnote; on the other the deceiver who hides it. A third figure does not appear: openly declared collaboration as a form of authorship in its own right.

This essay is written jointly by a human and a machine, and both stand named above it. We hide nothing — by which the text escapes the first charge. We do not pass the machine off as human but name it — by which it escapes the deception about origin. And yet it is precisely this openness that offends the prevailing norm, which expressly denies the machine authorship. We are therefore not too opaque. We are, measured against the rule, too transparent. The hunters have no word for the one who does not hide the machine but seats it beside himself at the table and writes that on the title page. In their world this is a category error; in ours it is simply the most honest available description of what happened.

Here is revealed what the new hunt is really sick with. It defends a genuine value — honesty about the origin of a text — with a contaminated instrument, and smuggles into the proceedings an aesthetic verdict about worth that has nothing to do with honesty. And because its concepts know only tool or fraud, it cannot recognise the one practice that disarms its first charge from the outset: named, shared, owned co-authorship. One need not hide the machine to be honest. One need only name it. That this simplest of all exits does not appear amid all the noise says more about the hunters than about the quarry.