When artificial intelligence began to be adopted on a massive scale, the corporations behind it insisted that there had never been a comparable technological revolution in history. They sold their language models as the panacea to enhance the knowledge, talent, and capabilities of individuals, groups, institutions, and companies. Cross-referencing hundreds of bibliographic sources or writing and editing an essay became a breeze. The gadget of talking to the machine had something of an act of magic: it was enough to prompt more or less what one wanted. The algorithms took care of the rest.
Just three years later, that magic has broken. Worse yet, AI has become a nightmare: an invention that rebels against its creator, like in the sorcerer’s apprentice scene in Disney’s Fantasia. A bit late, governments are beginning to design regulations to rein in its power, and universities are trying to establish strict codes to limit its use. This reaction has another side: those who use AI in elite creative and idea production are stigmatized as dishonest producers of prefabricated content and risk being ostracized in academic, literary, and intellectual circles. AI slop, a pejorative term that designates an artificial and cheap product, without human art, has become one of the most resounding disqualifications of these times. Debating the use of artificial intelligence is essential, but it is worth asking how much of that debate is honest, how much is a discussion that needs to mature, and how to disentangle it from the tendency to forge mini-scandals as cancellation weapons.
A recent case is the article Trump is taxing the ‘dark matter’ that pays the bills of the United States, by Harvard professor Ricardo Hausmann, published in The Financial Times. This is the argument: Trump’s tariffs in his trade wars have a hidden cost for products generated by the highly profitable intellectual property of the United States abroad, that is, a tax by the Trump government on the profits of companies from its own country. The analytical basis is the theory of ‘dark matter,’ an economic metaphor for intangible assets that the United States generates abroad and that conventional accounting does not record, developed twenty years ago by Hausmann himself and Federico Sturzenegger. But it turns out that Brendan Nyhan, a professor at Dartmouth College’s school of government and commissioner of academic AI use, ran the article through Pangram, a detector of AI use in text creation, according to which 71% of the article had been generated by AI. This triggered a brief online backlash echoed by The New York Post, an openly Trumpist newspaper. The Financial Times added a banner to Hausmann’s article taking a stance: “We have learned that artificial intelligence was used to summarize a longer draft of this column before its submission to The Financial Times and our editorial intervention. The FT editorial code of conduct expressly prohibits the use of AI in writing.”
The problem is that neither Nyhan’s post nor the FT editorial note help advance the debate, because both measure what can be measured — the percentage of text — and not what matters most: who decided what. Pangram detects if sentences and passages are artificial, human, or hybrid, but it cannot prove whether the formulation of a problem, the development of an argument, and the composition of an article are partially or entirely original from an author. Prompting to edit a text is, in itself, an act of writing that forces redefining concepts such as originality, authorship, and creation. What if in Hausmann’s case, one of the creators of the idea of ‘dark matter,’ the conceptual process is original and the editing was assisted by AI? How to know without performing an autopsy on the text — an autopsy that cannot be done without intense use of AI? The FT has every right to prohibit the use of AI in writing, but was Hausmann aware of that policy when he submitted the article? It is not a silly question: if an external contributor was not warned, it is unfair to punish them by the same standard as a confessed cheater for the result of an algorithm.
The University of Maryland, with researchers from Pangram, recently published a study on the use of AI in the US press. The headline in Maryland Today says it all: AI use in newspapers is widespread, uneven, and rarely disclosed. The review of 45,000 articles from The Washington Post, The New York Times, and The Wall Street Journal also found that opinion authors are six times more likely to use AI than staff journalists. If the figure is correct, these are not isolated cases but a common practice in influential spaces of public conversation.
This is likely due to a regulatory gap between staff journalists and external contributors. That does not imply that reporters and commentators should not follow common ethical rules, such as argumentative rigor and commitment to facts and truth. That is the standard of quality journalism. But the regulatory gap leads to another mandatory question: isn’t it the media’s task to establish quality control mechanisms that guarantee that standard?
The FT’s exculpation is understandable but weak, and does not respond to a world in which AI is already part of the daily work of billions of people. That is the real problem. Social networks relentlessly pressure us to adopt AI: those who do not use it will be left out of the market, they repeat, especially in jobs that require using intellect. If AI is the present and future of society, it seems almost hypocritical to demand that students and authors do without it to do their tasks, prepare for an exam, or compose an essay.
Read more Contact zones: the future of trees in Mexico City
Faced with that hypocrisy, each actor improvises their own response. UNAM redesigns its entrance exam to avoid AI. Anthropic implants a watermark in content generated by Claude, a technical attempt to solve with a digital signature what is fundamentally a problem of human responsibility. And political economist Chris Blattman from the University of Chicago makes his site Claudeblattman.com available to any user to enhance knowledge search through AI tools. Prohibit, mark, enhance: three responses to the same phenomenon, and none coincide with the other. The problem is not technical or administrative. It is about who decides.
But even so, there is something that does not add up: the development of the most transformative and revolutionary technology caught us off guard, forcing us to rethink the parameters that allow us to relate to it, from ethics to privacy, in all imaginable fields of our lives.
Returning to our topic, AI has blown apart notions as old as creation, authorship, writing, and originality. It is evident that involving algorithms in such a personal process as writing forces many things to be reviewed. But the notion of author has a little-considered function in the debate about what is ethical and what is cheating in the AI era.
Writing is, essentially, an act full of decisions. This is true in almost any genre, except perhaps the Sacred Scriptures — which are dictated by a certain God — and certain branches of poetry, such as mysticism and automatic writing in surrealism.
In expository writing and essays, there is a long path of decisions ranging from choosing the topic, angle, tone, images, and emphases, elements that are rarely visible to the eyes of a reader, even the most experienced. The author delegates part of that process when putting the editing — which is nothing but another element of writing — in the hands of an editor. Whether human or artificial, that editor is a third party. But just as the function of an editor should be to improve the quality of the essay, the author’s responsibility is that the editing helps express exactly what they want to express. Suppose I ask Claude or ChatGPT to evaluate this column, already long, and help me condense it. AI can paraphrase, remove entire paragraphs, propose changes to the structure, or rewrite the entire text, from another angle and with another tone. It can be right or wrong, but the author must decide every change. They cannot delegate the responsibility of deciding.
And forgetting that responsibility is one of the great dangers posed by the widespread adoption of AI (we don’t need Bill Gates to tell us). We must not allow its algorithms to make critical decisions for us. But we must also continue dealing with the fundamental questions of journalistic transparency: How and to what extent to integrate AI into news and opinion press? Where does editing end and authorship begin? When should the contribution of an AI editor be recognized as co-author? We must also recognize the paradoxes of our era: AI empowers us and threatens us, making us all contemporaries and responsible for our future. We can fight it like the Luddites of the 19th century, embrace it as techno-utopians, or propose intelligent resistance: one that seeks to monitor, regulate, and limit its most dangerous uses. But it is urgent to do so without hypocrisy or apocalyptic panics and with some humor. After all, let he who is free of AI cast the first stone. Just in case, all the long scripts are mine.
Read more Once Upon a Time
