The first distinction to be made is between the use of AI and the reproduction of copyright-protected texts. In France, using AI to write or revise a book is not, in itself, a criminal offence. A legal issue could arise if the generated text reproduced protected passages, or if elements had been reproduced without authorisation. It would then be necessary to establish precisely what had been borrowed, the extent of such borrowing, the rights concerned and the circumstances of publication. A result provided by an AI detector is not sufficient to prove plagiarism: it proves neither the origin of a passage nor the existence of a copy of a previous work.
Reports concerning older texts are therefore more likely to give rise to a copyright dispute than the mere suggestion of AI use. The Quebec press reports, in particular, that a column published in 2013 allegedly reproduced, with modifications, an article from the Philippine Star, and that other texts allegedly contained passages similar to previously published works. These remain allegations reported by the media: their legal classification would depend on a comparison of the texts and an examination of the rights held by their authors or publishers.
If a rights holder believed that their work had been reproduced without authorisation, they could seek redress in the courts. In Canada, the law provides, in particular, for damages and other civil remedies in the event of copyright infringement. In France, the infringement of a protected work may also result in criminal penalties, but this would require an offence to be established and legal proceedings to be brought. French law provides for up to three years’ imprisonment and a fine of 300,000 euros for copyright infringement; noting this penalty does not, of course, imply that it would apply to the author in this case.
As for AI, the debate could also concern contracts entered into with publishers, statements made when the manuscript was submitted, or the manner in which the book was presented to the public. However, one would need to be familiar with these documents and the commitments made to determine whether a breach of contract or deception could be invoked. At this stage, the mere suspicion of AI use is not, in itself, sufficient to conclude that an offence has been committed.
The legal risk will therefore depend primarily on evidence establishing the use of copyright-protected works, the identity of the individuals or companies holding the rights, and any decision to take legal action. As at 26 September, the publicly available information consulted refers to allegations, reactions from publishers and sanctions within the literary world, but does not mention any complaint or legal proceedings brought against Thélyson Orélien. He may therefore face serious professional and reputational consequences without, however, having been convicted by a court.
