Context Window 18
This edition covers OpenAI’s “freedom-focused” pitch to the Trump administration on AI policy, an FOI revealing Peter Kyle’s use of ChatGPT for policy advice, Wiley’s gold-standard author AI guidelines, Adam Hyde’s open-source EasyJournal and the wider rise of AI-assisted “vibe coding”, time spent with Ukrainian publishers and the threat of AI training on Russian propaganda, an HBR piece on responsible AI, Sakana’s AI-generated paper passing peer review, OpenAI’s new Responses API for building agents, and AI’s growing role in commissioning decisions across creative industries.
It’s been a busy week, with London Book Fair, followed by a guest lecture on AI and publishing at Cambridge University. I’m writing this on the train home with a depleted social battery but lots to think about. OpenAI has provided “freedom-focused” recommendations to a public consultation on President Trump’s AI Action Plan due this summer: unsurprisingly, that is freedom to train on an unrestricted basis. The arguments made are familiar ones on Fair Use and US competitiveness versus China, though OpenAI is also asking for AI to be regulated at a federal level rather than state laws. Left to the courts, these arguments might be expected to play out over a period of years. A Fair Use determination by executive fiat could short-circuit that process; it also raises important questions for publishers re the longer term value of potential licensing deals if developers can get access to data without paying in the near future. I’ve previously suggested that every publisher should consider the implications of training as Fair Use: I would now ask, what would the implications be if this came by July of this year? On this side of the water, technology secretary Peter Kyle’s ChatGPT use has been obtained through a Freedom of Information request and shows him using it for policy advice. John Wiley released its author guidelines on AI in LBF week: compared to everything else I’ve seen, these are the gold standard for author-facing documentation. It would be interesting to know whether Wiley provides its authors with equal clarity on how it uses AI internally. The brilliant Adam Hyde released EasyJournal, a completely free and open source journal publishing platform developed using AI tools. For a bit of inside-baseball on the creation of the platform, Adam has blogged about it here. There’s a good analysis of this kind of AI-assisted, solo development (or vibe coding) in this piece by investor Andrew Chen, who sees AI accelerating the process of industries becoming software-centric (for publishing, I’d reframe his “small/boring/slow-growth” segmentation as “established/niche”, but the underlying thesis strikes me as persuasive). I spent time at London Book Fair with Ukrainian publishers through the excellent SUPRR programme, which I commend to all publishing friends. One of the topics of conversation was recent research on Russian propaganda being pushed to AI models and then reflected in outputs. (Interestingly, the response wasn’t to reject AI outright but to discuss ways to counteract misinformation through corrective data.) My old business school marketing professor Caroline Wiertz has co-authored a new piece on responsible AI for HBR: my key takeaway was the disconnect between the 87% of managers who acknowledge the importance of responsible AI practices and only 15% feeling prepared to do so. The piece outlines practical steps for companies looking to close the gap between AI ethics awareness and real-world implementation. Japanese AI startup Sakana, working with scientists at the Universities of Oxford and British Columbia, claimed the first completely AI-generated scientific paper to pass a traditional human peer review, a significant watershed for AI in academic publishing. OpenAI launched a new Responses API for businesses looking to build AI agents. For larger publishers, this looks like a really promising set of capabilities. However, if you’re building AI on a smaller scale, HBR has some useful notes on building personal AI assistants. Outside books and journals, it’s interesting to see AI being used in other creative industries: an example is the acquisition of Jumpcut Media, which uses AI to support commissioning decisions in film (disclosure, they’re friends and clients, but it was great to have a drink with Gavin and Jeremy from Storywise this week, who are building something similar for books).
This was originally published in my email newsletter. To receive weekly updates on how AI is affecting the publishing industry, sign up here.