News
Guardian News and Media reports that scam fan letters, some mimicking real novelists, are targeting authors whose books trained Claude.
Reported by
Guardian News and MediaAnthropic agreed to pay $1.5bn to authors whose books it used, without permission, to train its AI model Claude. The case was brought by writers whose work was taken to build the very systems now used to generate fake fan mail that targets those same writers. As reported by Guardian News and Media, if you receive a gushing fan email asking for money to promote your work, report it and do not pay.
Last year, Anthropic agreed to pay authors $1.5bn in compensation for using their copyrighted work to train Claude. In July, a group of publishers filed a lawsuit against Google over its use of copyrighted material to train Gemini. Both cases establish, in money and in court filings, that these AI systems were built by ingesting the full text of working writers' books.
That is also why the scam emails now landing in writers' inboxes feel so convincing. The messages arrive from invented readers with names like Bella Monroe, Thalina Rose and Tricia Claxton, praising a novel's "deadpan genius" or a nonfiction book's "moral courage" in language specific enough to suggest the sender has actually read it. Some writers have received messages written in the voice of real authors, including fake fan letters apparently from Ottessa Moshfegh, Sally Rooney and Donna Tartt. In each case the email leads to the same request: a few hundred dollars to cover the cost of promoting the book at a city-wide book club or reading group.
The Anthropic settlement matters because it puts a figure and a legal finding behind something writers had only suspected. The model producing these scam emails was trained on their own sentences, so the praise it generates can sound more accurate than a stranger's actual opinion would. That is the mechanism the scam depends on: a system fluent in an author's own prose is used to flatter that same author into paying out.
The settlement does not stop the scam mail arriving. It changes what a writer receiving one of these emails now knows, that the tool behind it was built in part from legal claims their own industry brought and won. The correct response to the email itself has not changed: report it rather than reply.
Received something like this? Paste it into the Flggd message checker before you reply, click or pay.
