Jenny Williams lived in Ridgway for years, and her parents are there still. These days she writes from Aotearoa, New Zealand. In between she spent about eight and a half years inside big tech, at Google and then Microsoft, including a couple of years on an AI ethics team. It is not the résumé you expect on the back of a novel, and it runs straight through her new one.
House of Liars is what Williams calls a "slow burn literary psychological thriller." Six writers are brought to a remote island to compete for a million-dollar publishing prize. One of the writers is an AI. On the first day, the residency hands each writer an AI-generated version of the novel they came to write, to read or throw away. Williams calls it "this kind of tempting object."
Williams and KVNF's Brody Wilson discussed the temptations of AI in their discusion. Williams is clear that the anger toward AI in the creative world is earned. Her own first novel, The Atlas of Forgotten Places, turned up in the training data of large language models, unpaid. "Whether or not it was legal, whether or not it's termed legal, it's clearly unethical," she told KVNF. "These companies were taking lifetimes of labor from writers and artists without compensation." But she thinks the conversation leaves something out. "One thing that is not acknowledged enough," she said, "is really how tempting and seductive these technologies are."
Asked what is actually wrong, Williams gave four problems, not one: how the models were trained, those same models now competing with the writers whose work trained them, a near-total absence of disclosure to readers, and whether readers care. Amazon's self-publishing platform asks authors to declare whether they used AI, she noted, but does not pass that answer to buyers, and it is a single yes-or-no box. As she put it, "it's a binary that contains an enormous spectrum."
The episode also takes up what is changing. Article 50 of the European Union's AI Act took effect on August 2, requiring providers of generative AI systems to mark their output in a machine-readable format. Systems already on the market have until December 2 to comply. Anthropic, which makes Claude, has signed the associated European code of practice and published its approach: which is an invisible watermark in generated text, and a signed record on generated files. By Anthropic's own account, finding such a mark is "not fully conclusive." It says content may have been processed by Claude, not that Claude wrote it. Someone who used the tool to proofread carries the same mark as someone who had it write the whole book from a single prompt.
Williams is not persuaded there is much there. "I don't think there's a there there," she said of AI writing. "There's no sentience behind it. There's no lived experience." What she wants readers to trust is the opposite: "I write not because I have answers, but because I have questions."
