The AI accusations have been coming like sniper’s fire, with unseen character assassins spraying bullets and taking people’s livelihoods.

Nigerian writer Jerry Falade managed to escape with his life, but he lost what is believed to have been a $2m book deal for his work, Call Me, I’ll Hide the Body. In a devastating Substack, he explains just how deep the flaws run in AI software — in his case Pangram — for ‘detecting’ AI use.

The problem is that AI is ‘trained’ on existing writing, but nothing is explained. There is every likelihood that the cultural biases of the trainer will become the default of the system. This has certainly been identified in AI facial recognition systems. Misidentification is far more likely to happen to African and Asian people than to white ones. Software like Pangram appears to be doing likewise for non-western literature. 

As Falade explains: ‘… my language, Yorùbá — a highly lyrical language rich in proverbs, aphorisms, metaphors, parallelism, and “not X, but Y” constructions. Many of these features are now treated as AI artifacts.’

Even worse, this sort of detection software cannot compare anything beyond what it has been trained with. Falade gives this example: 

‘In Yorùbá, even an insult should display imagination. A blunt “You jerk!” or “Go to hell” would disappoint your Yorùbá ancestors. They might turn in their graves if you wasted an insult by making it ordinary. We say things like, Ẹ rí bí orí ẹ ṣe rí pẹlẹbẹ bí ìdí ìgbákọ (“See how flat your head is, like the bottom of a food scoop”). But place something along the lines of “See how flat your head is, like the bottom of a food scoop” in a manuscript, and an editor trained to fear every unfamiliar metaphor may swear on their life that some hallucinating machine wrote it. What is recognizable as humor within one culture becomes suspicious when read outside it.’

Accept the inevitable

At the moment, the onus is on writers to prove they are not using AI, and this is leading only to a heightened sense of panic among writers and confusion among publishers. The publishing world has barely entered the 20^th century, and unfortunately doesn’t exactly inspire confidence that it has the systems and structures to withstand a crisis of the 21^st. No one knows what to do.

The answer, to my mind, is very clear: Assume some people are going to write their books using AI tools and some are not going to admit it, either.

And then, going forward, be a discerning reader. Ideally, the publishing industry would also provide discerning editors, though at the moment there seem to be high-profile cancellations rather than rejections or queries at a far earlier stage. This was the case earlier in the spring, with Hachette UK pulling a novel titled Shy Girl after claims that the book by Black author Mia Ballard relied heavily on AI. This despite the book already being on the shelves in the United States and selling 1,800 copies.

A more recent AI finger-pointing case comes from France, where a Haitian novelist was this month accused of using AI, with his book pulled from a prestigious prize shortlist. The investigation into Thélyson Orélien was prompted by a right-wing columnist using a social media account anonymously. According to the New York Times report:

The controversy grew out of accusations from a little-followed, anonymous social media account, which a columnist at Le Figaro, a conservative daily newspaper, later revealed he partly ran … The columnist, Samuel Fitoussi, said in an interview that he had no political agenda in any of the accusations and ran the book through an A.I. detection tool, Pangram, because readers had flagged it. He said he thought an anonymous account would be trusted more than his column in a right-leaning newspaper.

Words as weapons

The more Pangram gets weaponised the more targets will come into its sights. So far (with distressing predictability) the highest profile cases have involved Black writers. Who after that? Anyone who has a dissenting opinion? Anyone who doesn’t ‘deserve’ their success? It is very unclear who is operating the levers of the AI plagiarism inquisition. Is it the media? Is it crazy eugenicists? Is it tech companies? Is it the software? 

Regardless, it’s clear there is a bully in the machine. The question is how to disarm it.

One way is to just accept that AI and its slop are part of the new normal. In my own practice as a writer, it’s the computer I now need to outsmart. It’s depressing — I’d rather feel I was in conversation with other writers, but that is not this moment. The onus is on me to tell stories that AI cannot. (Yet. Until it’s trained on more of my books. * SIGH* ). It’s better to accept that AI is here, than have the machine cancel you — or hound you to death.

Pangram (whose name, hmmm, bears a resemblance to the word Panopticon) is the last thing any creative person needs in this climate, which is to have their speech policed by this strange new fever-dream reality where a piece of all-seeing software trained on millions of books has the audacity to tell writers what they can and cannot write. Or even worse, tells writers that they are plagiarising … themselves.

Has anyone drawn the actual line that isn’t supposed to be crossed? I remember when people were aghast at the launch of Cliff Notes (‘Can’t students just read the actual book?’), spell-check on Word (‘Don’t they know how to spell?’), and Google translate (‘Too lazy to learn a language?’).

The capabilities of much larger AIs like ChatGPT and Claude are exponentially more powerful, and they have already inserted themselves into everyday life. In my own circles, I know people who use them to travel plan, do basic admin, and even vibe-code apps.

People who want to cos-play being a writer on Instagram are going to use AI to write ‘books’. Stressed or lazy or burnt-out writers are going to use it to cut corners or meet deadlines. Engaged, energetic writers who are truly curious about our AI future are going to see what it can do, and maybe use it in their work. AI is here, whether we like it or not, and whether we are in any way prepared for the economic and environmental havoc the huge data centres are going to bring (to say nothing of the potential for a more general cognitive collapse).

None of this is to say that I support these developments, but there is no wishing them away. We have to learn to live with them, and the first step in doing so is to understand that these tools can and will be weaponised. So at least in terms of trying to neutralise that weapon in the sphere of books and publishing, the clearest path is to accept them and then try to find a way around them. 

The second is, by all means, to have the publishing industry or writers use some sort of disclaimer, along the lines of ‘written without the help of AI’. Like an organic cheese rating, but for books. The publishing world is having a full-blown crisis about the creeping use of AI, and yet many other dubious practices have long been accepted. For instance, the king of the airport thriller, James Patterson, employs an army of writers who publish under his name and he sells books by the truckload. I can’t compete with that. I don’t want to compete with that.

Ditto AI. Some people are going to use it and quite possibly lie about using it. Fine. So be it. The onus will be, as it already has long been, on the people who care about craft to keep improving their work until it’s quite obvious what is computer-aided and what is not. Writers will have to tell stories better than the computer. 

There is a very unscientific adage that ‘you are the average of the five people closest to you’. I’d tweak that to say that writers — and readers — are perhaps the average of the five books they are reading, or maybe their five favourite books. If you read slop, you’ll write slop. If you read quality, you’ll reach for excellence. AI is frightening and demoralising in the world of publishing, but the real hope lies in the very discernment that makes us human. Computers may say no, but humans — at least right now — can still say yes.

+++++++++

What I’m … doing: seeing Seoul as a tourist

I’ve not been at my desk much this past fortnight, between a terrible cold and then the arrival of friends from the UK. I’m hardly an old Seoul hand and there is plenty for me to learn, but a trip to Seoul Tower was a useful reminder of the value of pushing past your comfort zones. We live near this iconic tower, and I’ve taken more than half a dozen friends to its base, which is on Namsan mountain. The views from there are spectacular and, for the most part, sufficient. But it’s possible to go up into the tower to a viewing floor with even higher vistas. None of my guests had wanted to bother with that thus far. My husband and I had discussed going to the restaurant at the very top, but we still haven’t gotten around to it. So I owe Nick and Beth a big thanks for their enthusiasm in wanting to go as high as possible. It was almost as if there were a whole other city I was missing. Those extra 230 or so metres provide a different perspective, not least in that for the first time I could look down and see the entirety of the city below me and not just off in the distance. Also, the viewing floor has what might be the coolest loos in the world. To think I have been here almost three years and missed those! Travel broadens the mind, we’re told, but so does taking a trip in your own back yard.

++++++++++++++++

If you liked this essay, leave a comment. Or buy me a drink! Or treat yourself to one of my books!