Context Window 69
For authors, publishers and technology companies, it’s been a week that challenges the old axiom that there’s no such thing as bad publicity. The practical uses highlighted in the rest of this week’s newsletter need to be balanced against good strategy and judgement.* The news this week has been dominated by Hachette’s withdrawal of a hyped book after a New York Times piece questioned whether AI had been used in the writing process. This is a minefield for authors and publishers, and the more I read about it, the less comfortable I am.
There is a key point missing from the Times coverage which appeared elsewhere this week: evidence that particular types of prose, and writing from neurodiverse authors or those working in a second language, are more likely to be flagged or challenged by AI models. I’m not suggesting that was a factor in the Mia Ballard case. But I would be interested to know how AI detection companies respond to that concern. (I’ve asked that question of one of the leading companies, and I’ll update if and when they reply). However, if this is a possible factor, then anyone levelling accusations or taking action based on detection software needs to be alert to the potential for uneven or discriminatory harm.
This isn’t just an issue for trade publishers. Alexander Kustov recently posted a follow-up to his Substack piece on the impact of AI on academia, as part of which he concludes that AI detection software and disclosure norms simply do not work. The whole piece is worth reading.
Related, Nature reports that a leading AI conference rejected 2% of submissions after their authors broke the conference AI policy by using LLMs for reciprocal peer review. They weren’t caught through detection software, but through hidden, watermarked instructions that inserted specific, revealing phrases into the review text.
On the wider question of AI and writing, I have an article in the Spring 2026 edition of the Society of Authors journal, looking at how authors might think about a strategy for AI. My aim was to provide a calm, strategic view of an issue that too often swings between hype and panic—I’d love to know what you think.
The big tech news this week was OpenAI closing its video generation tool Sora. In the process, they burned their billion-dollar partnership with Disney, which reportedly learned of the change less than an hour after Disney and OpenAI teams had met to discuss their collaboration.
There’s a lot to unpack here: Sora was burning money (even by AI industry standards), carried a significant opportunity cost as well as a P&L one, and was likely a content moderation nightmare. More broadly, the closure reflects alarm within OpenAI at being outpaced by Anthropic. But there’s a fine line between a pivot and a lurch, and this will certainly give any future strategic partners pause.
It’s not all plain sailing for Anthropic though, which is facing a new copyright infringement lawsuit from music publisher BMG Rights Management (owned by Bertelsmann, Penguin Random House’s parent company).
A really useful practical update for publishers: Google NotebookLM has added support for EPUB files as a source type. This will be particularly useful for generating ancillary content like quizzes, infographics and videos from existing books. Of course, there are two groups of people with large collections of DRM-free EPUB files on hand—publishers are one, but what the other might choose to do with generative tools like this raises some obvious questions around rights and control.
WordPress, which powers over 40% of all websites, announced significant improvements to its support for AI agents and MCP, including the ability for agents to create pages, write and organise content, and update metadata (including fixing missing alt text). This potentially addresses pain points that I regularly hear about from publishing marketing and web teams.
Does anyone know a data-oriented book lover looking for a new job? ElevenLabs is hiring a content curator for its ElevenReader audiobooks product.
The Higher Education Policy Institute has published its 2026 survey on student use of generative AI, in association with Kortext. Last year’s usage numbers were so high that there was only so much further they could grow. This year’s findings included 94% of students using AI to help with assessed work, 68% seeing AI as an essential skill, but under 50% reporting that their teachers helped them to develop appropriate skills. It’s essential context for anyone interested in the impact of AI in education.
This was originally published in my email newsletter. To receive weekly updates on how AI is affecting the publishing industry, sign up here.