Finally, one of the more buried nuggets in the interview are three insights, the first being, that digital technology – including AI but not just – has really enabled a form of investigatory work that was very difficult to do at scale and speed previously.

While Brown points to the challenges of AI for investigations, it’s also incredibly interesting to see how he elaborates on some of the possibilities that come from the use of AI for the exact work he and colleagues the world over are doing.

Brown’s emphasis on showing their process (& hope that audiences will see this as a sign of quality) fits what we’ve termed the provenance premium: that reliable and transparent actors become more important as credibility anchors and will be more trusted, at least by some

Brown highlights the increasing uncertainty over what is real and the challenge to the indexicality of images and audio as a result of very capable GenAI tools.

Slightly befuddled by Google doing this but equally irritated by claims that this is going to be a disaster, misinformation wise: x.com/Google/statu... Some thoughts below…

On the labour creation effects of a new technology, here Gutenberg’s printing press: “Finally, printing required more workers and more division of labor than scribing” (from Ann Blair’s magisterial “Information in Early Modern Europe).

But also: "Given these risks, various participants argued for better independent evaluation and testing but noted that transparency around model evaluation and leaderboard claims (which are currently being wildly used to assess model and agentic capabilities) is thin"

"Interaction risks also compound at a system level. Panellists warned that components that look safe in isolation may, when networked, amplify vulnerabilities: multiagent configurations can jailbreak each other; data-gathering agents may breach implicit bounds"

Last summer, I co-organised a small symposium on AI agents which in light of the OpenAI/Hugging face news feels prescient. Some quotes from the symposium report: "Several of those present argued that many of the most pressing risks are systemic."

In September I'll be spending three weeks at the Institute for Advanced Study of the Research Alliance Ruhr – the @college-uaruhr.bsky.social in Essen – as an Invited Fellow, by kind invitation of Julika Griem. Would be great to meet people in the area while I’m there.

For four years, AI has reigned supreme – and lots of news organisations have jumped on the bandwagon/caught on/thought deeply about how they, too, could make use of it. What do they (want to) use it for? Increasingly for content creation. Full post here: open.substack.com/pub/felixsim...

AI coverage 1, Climate coverage 0: In comparison to AI, climate change barely received any media coverage as of late. open.substack.com/pub/felixsim...

Fuller piece here: substack.com/@felixsimon/...

8) 42% say they always/often click through to original sources, but lower than for search/social/video; but recall bias! Those who do click through are motivated more by verification than by wanting more detail 9) Concern about fake news up 4pp to 62%

7) Trust in AI chatbot news answers low: 20% globally vs 37% trust in news overall; UK lowest at 6%

3) Higher adoption where search/social/aggregator/video use is already high; lower where direct access dominates 4) Top use: follow-up questions (42%), latest news (35%), summarisation (34%), easier-to-understand news (30%).

So what are the main DNR AI findings? 1) AI chatbot use for news up to 10% (from 7%), but only 1% cite it as main news source 2) Usage skews toward news-interested, AI-trusting, and younger audiences (16% of under-35s) South Korea, Spain doubled YoY; USA, UK, Germany flat

5) The UK public mostly seems to care about health, immigration, the economy/cost of living if you trust YouGov and ONS polling.

3) UK media generally covered AI generally on par with topics like immigration, climate change, cost of living/the economy, but somewhat below health and the NHS.

1) It's not true that leading UK media did not cover the story. 2) It's true that it was not top of the list (but it wasn't exactly buried either)

Against my better judgment, I wrote something short on the claim that UK news isn't paying enough attention to the Claude Fable quagmire and AI more broadly.

✨ 🤖 New pre-print: Is realism dead and do people distrust everything, now that synthetic media is ubiquitous? It’s a bit more complicated.

Not entirely sure I agree with everything argued in this short commentary pre-print on AI slop, but I do think the authors get one thing right: It’s too easy to dismiss AI slop.

Today in something that might not be entirely surprising but is still worth stating clearly: When we ask people to think about the what AI will do to news and their experience of it, their answers do not emerge from a vaccuum of perfectly rational and neutral observation.

Interesting work by the New York Times' Ted Alcorn, letting you dig into publication trends at the the NYT between 2000–26. Notable that the paper publishes significantly fewer words per month (and fewer articles) than 20 years ago. tedalcorn.github.io/nyt/

I am happy to announce that @politicsoxford.bsky.social has awarded me a grant to study how agentic AI systems can be better aligned with democratic values in the domain of news and public-affairs information.

At the @journalismfestival.com next week? If so, come and join me, @marinaadami.bsky.social and Agnes Stenbom if you want to talk about AI ✨🤖 Details below. (And hit me up if you want to get a coffee or aperitivo to chat some more in Perugia).

Speaking of German military ships, it will never not be funny that all four ships of the F125 frigates – notorious because the first ship was overweight & had a persistent 1.3° list to starboard so basically swum in circles – are all named after landlocked federal states. Peak German humour.

Nominally an article about Germany’s military procurement issues, this paragraph speaks to a deeper truth about the country – antiquated practices that drive people (including many Germans) up the wall. www.ft.com/content/124c...