Sarah Knafo’s case is particularly interesting because she makes no secret of the fact that she used AI to write *The Heist of the Century*, published by Fayard on 2 September 2026. According to the publisher’s details, the book runs to 286 pages and is presented as a wide-ranging investigation focusing in particular on public spending.
When asked directly by *Le Point* whether she had “used AI to write this book?”, Sarah Knafo gave a much more nuanced answer than a simple yes or no.
She explains that she used Perplexity to verify figures and sources. She also states that she used Codex and Claude Code to create tax-cut simulators. However, she insists she did not use artificial intelligence to write her book, explaining that she enjoys writing too much to do so.
We must therefore be precise: there is currently no evidence to suggest that AI wrote *The Heist of the Century.
But this is precisely where the story becomes interesting.
For what does it mean to write a book?
Is it merely a matter of producing the sentences that will ultimately appear on the page?
For centuries, the work of an essayist has also involved researching documents, cross-referencing information, verifying figures, carrying out calculations, formulating hypotheses and then organising this material before putting it into writing.
However, part of this work can now be significantly accelerated by artificial intelligence.
In Sarah Knafo’s case, therefore, the AI did not actually do the writing. But, according to her own statements, it did contribute to certain stages prior to the actual writing.
This distinction is important.
Using a calculator to check a percentage obviously does not mean that a calculator wrote your book. Nor does using a search engine. But the new AI tools go much further: they can search for, summarise and compare information, generate code and build tools for exploring data.
The line between a ‘research tool’ and an ‘intellectual collaborator’ is therefore becoming much less clear-cut.
And this issue extends far beyond Sarah Knafo.
In a few years’ time, asking a writer ‘Did you use AI?’ may seem just as strange as asking them today whether they used Google, Wikipedia, Excel or a spell-checker.
The real question might become: What did you ask the AI to do?
Find a source? Check a figure? Summarise a hundred reports? Build a simulation? Suggest an outline? Rewrite a sentence? Write an entire chapter?
These are obviously not the same uses.
Sarah Knafo, incidentally, provides some interesting insight here by specifying precisely which tools were used and their functions. This allows the reader to distinguish between the work claimed by the author and that entrusted to machines.
Her use of artificial intelligence does not, moreover, stop at the preparatory work for the book. Her communications regarding *The Heist of the Century* have also publicly utilised content generated with the help of AI.
But this still does not prove that her text was written by a machine.
And this is perhaps precisely where the upcoming debate lies.
We are probably among the last generations to believe that there is a perfectly clear line between writing alone and writing with a machine.
In the future, this line will consist of dozens of nuances.
*The Heist of the Century* ultimately offers a rather revealing example of this new era: a book whose author fully claims credit for the writing, but in which some of the research and technical work has benefited from artificial intelligence.
It is not a book that we can say was written by AI.
It is something far more contemporary, and probably far more interesting:
a book written by a human who is already working in a world where artificial intelligence forms part of the author’s toolkit.
