Rachel Coldicutt

This a classic case study for anyone working in tech policy www.anthropic.com/news/fable-m.... I don't pretend to know anything about international trade law but there is a bunch of interesting stuff here. 🧵

Statement on the US government directive to suspend access to Fable 5 and Mythos 5 anthropic.com

Hmmm. I have been under a rock so missed this, but Nvidia buying Hugging Face does *not* seem like great news blogs.nvidia.com/blog/nvidia-...

NVIDIA to Acquire Hugging Face blogs.nvidia.com

Hang on, there's another "AI is terrible" letter? And this one is signed by Jack Clark and a bunch of other Anthropic/OpenAI folk. I hate to say this lads, but I have news for you... www.nytimes.com/2026/07/13/b...

nytimes.com

The AI Resist List is here airesistlist.org - so much hope and positivity for crafting positive, different futures in the room and on the list.

The AI Resist List airesistlist.org

Looks like this not off to a flying start www.politico.com/news/2026/09...

politico.com

I am such an old-fashioned stick in the mud because perhaps the answer is not to (a) integrate these models into every frickin' thing and (b) stop turning the people who make them into unstoppable gods www.theguardian.com/technology/2...

AI in your oven? Anthropic warns models ‘months away’ from self-replication on smart devices theguardian.com

I do find this whole thing a bit weird. Anthropic could, literally, stop or change what they are doing at any time www.theguardian.com/technology/2...

Anthropic says the world should have option to ‘pause’ on AI theguardian.com

"They are arguably the modern-day railroad,” Schulte said. “It makes sense for them to issue longer maturities because AI is a long-term infrastructure." giftarticle.ft.com/giftarticle/... I'm v sympathetic to long-term infra development, but what if hyperscale AI is as flying…

ft.com

This interview with Simon Johnson, chair of the new UK AI Economics Institute, is interesting not so much because it is very revelatory but because his views will presumably shape policy and ministerial opinions giftarticle.ft.com/giftarticle/...

ft.com

I just noticed that the website for the UK Gov Sovereign AI Fund looks a bit like it's promoting either a video game or a new Star Wars movie www.sovereignai.gov.uk

UK Sovereign AI Fund | Backing British AI Startups sovereignai.gov.uk

Oh, also - in democracy-related uses of AI - this OpenAI cheatsheet for policymakers is 👀 (also interesting that they couldn't render the prompts in a fluid responsive way - I have c+p'd one so you don't have to) academy.openai.com/public/resou...

OpenAI Academy London - Resource | OpenAI Academy academy.openai.com

Just opened Miro for the first time in a little while to find that it won't let me just draw a diagram, it wants to me to hang out with its little chatbot friend. So I ask it to draw three concentric circles.

Okay, fellow AI/sociotechnical researchers, I'm looking for a name for something but can't find anything in my lit searches: is there a term for the phenomenon where someone has false confidence in an AI output? ie a developer thinks their code is more secure, but it's actually less secure

This is a fun time to work in AI if you are on multiple people's list of "emergency women to invite to an event when you remember that not everyone in the world is a man under 40"

Someone had just emailed me on a Friday night to ask if I have any thoughts to share on AI in digital govt and I think my answer is that we should focus on making basic services work for everyone and stop worrying about AI

Spending this warm July evening making a list of AI CEOs who have signed open letters fretting about the risks of AI and then continued to develop AI.

I've done the same "understanding the consequences of AI" talk twice today, and sometimes it's still a surprise even to me just how wide-ranging and varied the harmful impacts of technologies can be when they are badly made and rolled out without care and attention.

I have an evolving and slightly problematic observation about LLMs and chat bots which fits with none of my previous opinions, so why don't I try it out in a low stakes way on Bluesky where absolutely no one will take exception and shout at me. What's the worst etc

I've spent the last year working on a review of the safe adoption of AI in safety critical systems, and I think one of the hardest things to communicate is that probably Sam Altman's biggest flex has not been in any invention but in disrupting the "traditional" model of innovation diffusion

I think talking about the scope creep of AI development as being about power and control turns a lot of people off, so what if we think of it as greed? Pac-man bots eating up all the knowledge and joy in the world till 3 or 4 companies have it all and then use it to autocomplete marketing emails

One of the most interesting things that came up on 5Live just now was someone who rang in to say, quite reasonably, and I've heard a lot of other people say this too, "I can get a lot out of working with LLMs because I'm an expert in my area, and I know at a glance when they go wrong".