Context Window 85
It’s been hard to ignore significant legal and policy developments for AI and publishing in recent days, with a major new class action lawsuit in the US, policy developments in Australia and regulatory jeopardy for Google and Perplexity in Germany. However, if you’re more interested in things you can do today, there is some excellent practical advice on developing policy and preparing your website for AI search.
The biggest news of the week was a new class action lawsuit for copyright infringement filed by publishers Hachette and Cengage, and author Scott Turow, against Google. If you’re experiencing déjà vu, the same plaintiffs sued Meta and Mark Zuckerberg in May, and joined a separate, author-led class action against Google in January.
As I posted in a Bluesky thread earlier in the week, there are some interesting aspects to this latest case. The plaintiffs seem to have taken on board Judge Chhabria’s comments in Kadrey v. Meta and have set out a really clear argument for how Google’s AI training has harmed the class. Authors and publishers will also be watching this closely in case of another settlement like Bartz v. Anthropic (for those outside the US, once again US copyright registration is necessary for class eligibility).
Germany’s media regulator ruled this week that AI summaries generated by Google and Perplexity are content published by the platforms as opposed to third-party information, making them subject to regulation. Earlier in the year, a Munich court found Google liable for the content of summaries. This creates a new area of legal jeopardy for AI platforms. The interesting question is whether this decision creates a precedent with other jurisdictions following the German lead, or whether AI companies could decide to reduce availability of AI features where there is risk (as with Meta’s delayed release of its AI models in Europe).
On a more practical note, Faber’s Head of D2C Marketing Tim Woodall has written a guest post for the Cultural Content newsletter on how brands should respond to Google’s AI mode and shifts in consumer behaviour: it includes practical steps that any website owner could take.
CJR has a good piece on how to develop an AI policy for newsrooms, which would be broadly applicable to other types of publisher. It offers sound advice and perspectives from a range of organisations.
Last year two sets of research from BISG/BookNet Canada in North America and the IPG in the UK suggested barely a third of publishers had a policy. I’m sure you all do by now, but maybe pass this link to your friend who still needs one.
The kicker, of course, is that both pieces of research were conducted by the same company, albeit with different survey designs and participants. One important difference is that the first study used a controlled experiment and the APA one asked general opinions. The way you frame and ask the question shapes the answers you get. And what people say and what people actually do are often different things.
Anthropic released a new feature, Reflect, which summarises how users have interacted with Claude—a bit like Spotify Wrapped for an LLM. Besides the novelty value, I think there’s something much more interesting going on here: in my training work and my forthcoming book, I talk about the importance of metacognition, being reflective about how you use LLMs and how it impacts your thinking. That is one of the best ways to improve your usage and guard against over-reliance on AI. I hope that other models introduce similar functionality.
A new analysis suggests that OpenAI’s plans to monetise ChatGPT through advertising in outputs may miss the company’s forecasts by over 90%. There’s an obvious caveat that these are long-range predictions and a lot could change, but if this revenue stream does disappoint it could point to more aggressive pricing to customers in future.