We have to care, thanks to the Luddite Fringe McCarthyist tactics and AIs-under-the-bed scaremongering. Because hiding behind the next bill board is a Luddite AI-catcher cop with an AI detector set to “Every word is AI – let the bastards prove otherwise.”
In the space of eight days as October faded, the UK’s Society of Authors (SoA) launched an “Ethical AI Charter”, declared that ethical use of generative AI is impossible, and then withdrew from the charter it had just launched.
Meanwhile, in France, a group with no standing in the literary establishment announced a prize for books written with AI, and in doing so asked the question the SoA has spent three years refusing to ask: what if we simply judged the work?
Yeah, I know, an outrageous idea. Judging content on merit? A slippery slope.
What the French have done
The Prix Pascal opens for submissions on 2 November. French-language writers, professional or not, can submit a manuscript of 10,000 to 50,000 words, together with the prompt that started it (100 words at most), a technical account of the model and process, and a formal declaration that AI was involved.
The method has to be described in enough detail that another writer could reproduce it. A five-member jury, reportedly to be chaired by Jacques Attali with CNRS researcher Alexandre Gefen among its members, will pick ten nominees from entries closing at the end of November, with a winner in December.
The timing is no accident. Days earlier, Thélyson Orélien’s novel was pulled from the Goncourt race after detection software flagged it, and his publisher later confirmed AI had been used. That is the approach French publishing has settled on: run the text through software, argue about the score, pillory the author. Sound familiar?
The Pascal organisers propose the opposite. Make disclosure the price of entry, and judge what is on the page.
Back when Rie Kudan won Japan’s Akutagawa Prize in January 2024, and then revealed she had used ChatGPT for around five per cent of the text, TNPS argued that the debate about whether AI can match human quality had just been settled by a panel of human judges who called the book “almost flawless” before they knew.
The lesson from Japan was that blind judging of AI-assisted work is a trap for everyone: the prize, the author and the reader. The Prix Pascal is the first prize I am aware of to act on that lesson, with disclosure up front rather than as a confession afterwards.
The French organisers even list encouraging the use of AI among their stated aims. That is refreshing candour. Most institutions in this business pretend to neutrality while behaving as prohibitionists to keep the baying wolves of the Luddite Fringe at arm’s length.
Fair questions
None of this is settled, it should be noted. At the time of the announcement the prize had no confirmed jury and no declared prize money, and its founders come from business and technology rather than letters.
Gefen himself has called it a playful experiment. And it gets weirder. The requirement that entries address ecological, social or ethical challenges is an odd one for a literary prize, and nudges the field towards message fiction. And “reproducible” is a stretch for systems that never give the same answer twice.
These are the weaknesses of a new project, but they are fixable. They are also not the point. The point is simply that somebody is running the experiment at all.
At which point let me spin off at a slight tangent, because as I wrote those words, I thought, does that sound too much like AI-speak? So I asked AI. (Google AI, because I had the Google window open.)
What follows is, verbatim, what Google AI said when I ran that paragraph past it. (“These are the weaknesses of a new project, and they are fixable. They are also not the point. The point is that somebody is running the experiment at all.”)
Yes, it sounds exactly like what a human would say to defend a messy but exciting project—but it relies on a specific philosophical concept that AI heavily uses.
Your phrasing doesn’t sound “robotic” or like standard ChatGPT-generated corporate speak. It is punchy and direct. However, it mirrors AI discourse because you are framing the project through the lens of an “existential proof of concept.”
Here is a breakdown of why it hits that fine line, and how you can tweak it depending on your goal.
Why it feels human
- The tone is dismissive and pragmatic: AI writing is usually overly polite, hyper-structured, and eager to please. Your phrasing is blunt: “They are also not the point.” That abrupt pivot feels like a founder or developer talking.
- It focuses on agency: Believing that “somebody running the experiment” matters more than the immediate flaws is a deeply human, entrepreneurial mindset.
Why it might trigger an “AI radar”
- It treats execution as the sole metric: Generative AI has flooded the tech and creative worlds with “experiments” that look cool but are broken under the hood. Tech commentators frequently use this exact defense for AI tools: “Yes, it hallucinates, but the point is we are witnessing the experiment live.”
- It uses “AI buzzwords” implicitly: Words like weaknesses, fixable, point, and experiment arranged in short, definitive declarations match the stylistic pattern of AI-generated summaries or LinkedIn “thought leader” templates.
How to tweak it
If you want to ensure it sounds 100% human and less like a tech-bro or an LLM, you just need to inject a little more personality, specific intent, or conversational texture.
- To make it sound more casual/gritty (More Human): “Look, the project has flaws, but they’re easily fixable and frankly beside the point right now. What matters is that we’re actually out here building it and running the experiment.”
- To make it sound authoritative/visionary: “The technical weaknesses are fixable details. The real value is the momentum—the fact that someone is actively driving this experiment forward.”
- To lean into the tech-forward style (if that’s your goal): Leave it as is. It communicates a bias toward action over perfection, which works well in pitch decks or startup environments.
Thanks, Google AI.
Personally I would have responded, “Who the f**k cares?”
But of course, nowadays we all care. We have to care, thanks to the Luddite Fringe McCarthyist tactics and AIs-under-the-bed scaremongering. Because hiding behind the next bill board is a Luddite AI-catcher cop with an AI detector set to “Every word is AI – let the bastards prove otherwise.”
Meanwhile, in London
On 13 October the Society of Authors, with partner bodies in the US, Australia, Canada and New Zealand, launched a charter under which agents, publishers and prize bodies could pledge to use generative AI in a transparent, responsible and ethical way.
The Charter was backed by literary notables like Kazuo Ishiguro, Tracy Chevalier and Kester Brewin, so of course the SoA jumped in with both feet.
Explained Ganley: “The Society of Authors has joined forces with international writers’ organisations and other creative industry bodies to launch a global charter setting out commitments to the responsible and ethical use of generative AI services.”
Got that? “Setting out commitments to the responsible and ethical use of generative AI services.”
In case we were still were unclear, Ganley added, “While this charter won’t solve the large-scale societal threats that unregulated generative AI poses, it is a small step towards defining what ethical and responsible AI might look like in the creative industries.” The Bookseller carries all the juicy bits.
Given the SoA is so vehemently against AI that is barely capable of rational discourse (except if its Amazon AI, which gets a free pass) –
I wondered how the SoA had ever backed the charter in the fist place. I wasn’t alone.
Three days later, under fire from members with lots of social media followers, chief executive and candle in the wind Anna Ganley was walking back every word. “There can never be any truly ethical use of generative AI.”
On 21 October the SoA withdrew from its own charter, promised a member survey, and parked the question for its AGM on 19 November. In other words, Ganley is passing the buck to the AGM. You couldn’t make it up.
Ganley nonchalantly explained, “The Society of Authors’ position on generative AI remains unchanged.” (The Bookseller.)
Consider a moment what that sequence says. A charter defining how ethical use might look, backed by reputable writers’ organisations in the US, Australia, Canada and New Zealand. Free publicity! A photo-op to to sound relevant! Everyone aboard! And then a reality check. A CEO saying there is no such thing. A withdrawal. A grovelling apology.
I guess we should be grateful for the apology, at least. But given the looming AGM, what else could Ganley do? Normally when the SoA gets it wrong, it just looks the other way.
Anyone remember (don’t worry – I won’t let you forget) when the SoA stormed into the news that Spotify was going to carry audiobooks, falsely claiming the Swedish company was giving away audiobooks for free and the end of civilisation was nigh.
But back to the latest charter the SoA jumped into and jumped out of again, without actually thinking either time.
The sharpest criticism I have seen, from the author KJ Charles, was that the charter’s very existence concedes ethical use is possible. The SoA managed to be too absolutist for the real world and too accommodating for its own members, at the same time.
And now Ganley, again no thought involved, just knee-jerk reaction to a headline, has stated categorically that “there can never be any truly ethical use of generative AI.”
Never, I tell you. Never!
But “never” is a strong word to paint yourself into a corner with. This is from the same SoA that insists AI companies must do deals with authors and publishers. If no ethical use is possible, what are the deals for? How can they even happen?
The defensible answer, of course, is compensation for harm done. But a body that argues that cannot also write a charter on how to do it properly, and cannot expect anyone to take either position seriously.
The letters (oh so many f*****g letters!)
In the same breath, Ganley said the SoA is supporting SoA Australia by writing letters to Australia’s attorney general and prime minister about the AU government’s AI copyright proposals and how they will affect international authors. Canberra is considering an opt-out scheme, and a statutory or extended collective licence, for AI training on material that is “unprotected” on the internet.
Whether that design is right is for Australians to argue. The Australian Society of Authors is perfectly entitled to lobby its own government.
But forgive me if I find it breathtakingly arrogant that a UK trade union, one that could not hold a charter together for ten days, thinks the government of Australia should rethink its policy because the UK’s Society of Authors doesn’t like what it is doing.
The tone is that of an organisation that has mistaken its own anxieties for a global mandate. Which is exactly what it is. I’m sure the Australian govt is quaking in it boots worrying that Anna Ganley might be coming after them.
As we all know, Anna Ganley regularly fires off epistles to Donald Trump explaining why the US position on AI is wrong. Except, of course, she doesn’t. The letter to the AU PM is performance, nothing more, and that’s what the SoA does best.
But hey, the SoA is a master of the art of a pithy performance letter, and knows all the best photo-op postboxes.
Back in 2024 the SoA wrote a stern letter to all the “tech bros” (except Amazon) that took the SoA all of one hundred and eleven days to write. Needless to say the AI companies ignored it.
The Meta Letter: Performance art masquerading as advocacy
The Society of Authors’ approach to AI advocacy perfectly encapsulates its fundamental unseriousness. Consider the organisation’s much-publicised letter to Meta protesting the use of copyrighted works in AI training.
The SoA sent this letter by snail mail. Not email. Snail mail. Why? Because posting a letter through a physical letterbox makes for a better photo opportunity than clicking “send” on an email. The performative gesture mattered more than effective communication. A masterfully meaningless performance. Needless to say Meta did not respond.
The numbers game
Another exhibit in the long-running “Look at us – we care so much” saga is Brave New World? Justice for Creators in the Age of Gen AI, published on 30 January by the SoA and four other creative bodies (the Independent Society of Musicians, the Association of Illustrators, Equity and the Association of Photographers), drawing on surveys of more than 10,000 creators gathered between 2022 and 2025.
The headlines for authors: 86 per cent say their earnings have already fallen, 72 per cent say job opportunities have been cut as clients turn to AI, and 57 per cent say their careers are no longer sustainable.
Classic SoA headlines. Classic SoA bullshit. Set those figures beside the last independent measurement of what UK authors actually earn. The ALCS-commissioned CREATe survey, with fieldwork beginning in February 2022, nine months before ChatGPT launched, found that the median earnings of primary-occupation authors had fallen from £17,608 in 2006 (in 2021 pounds) to £7,000, a drop of around 60 per cent. The share earning all of their income from writing had fallen from 40 per cent to 19 per cent.
The survey’s lead researcher put the accelerated decline down to world events: the pandemic, Brexit and the cost-of-living crisis. It is no gold standard, with 2,759 responses from 58,260 members invited, but it asked authors what they earned. (Brave New World or Same Old Bulls**t? The UK’s Society of Authors Again Shows It Is Not Fit For Purpose.)
That is the distinction the SoA deliberately blurs. The ALCS survey measured earnings. Brave New World asks authors to attribute them, after years of being told by their own union that AI is the cause. A decline that began in 2006 may well have been made worse by AI. But a survey that invites respondents to name the culprit cannot separate a new cause from an old trend. The test is not difficult: compare the post-2022 trajectory with the pre-2022 trend.
Meanwhile publishers have been signing AI licensing deals, with authors at houses such as Taylor & Francis complaining of neither opt-out nor payment. Those are contract problems, and the SoA could attack them today without waiting for any government. Whether the SoA leads this panic or merely follows its members into it, a union that cannot tell its members the difference between a trend and a cause is failing them.
Which brings us back to October 2026 and the SoA and a French initiative, each at opposite ends of the common sense spectrum.
Two models
One model says: disclose, document, judge the work on its merits, and accept you will sometimes be wrong.
The other says: declare the whole enterprise unethical, demand payment, write to foreign governments, and withdraw when your own members dislike your own charter.
The French prize takes its name from Blaise Pascal, the writer and mathematician who invented the Pascaline calculating machine. A writer who built a machine. The French, at least, remember the two were never opposites.
This post first appeared in the TNPS LinkedIn Analysis newsletter.
