Claude comes to Whitehall
Inside Anthropic’s first big UK policy event
If you want to make a statement with your first big event, it’s hard to imagine a better setting than Banqueting House.
With its Palladian structure and fantastical painted ceiling, the last remaining vestige of the Palace of Whitehall made for a very grand backdrop to Anthropic’s first major UK public policy event last week. It also happens to be the site of Charles I’s execution, but we won’t read too much into that.
The morning brought together researchers, people from government and think tanks, and others from the AI ecosystem. This was not like the hacker-focused industry events that have been driving hype in recent weeks. Instead, it served more as an official introduction to the company; a Westminster debutante ball for Claude.
Anthropic now employs 200 people in the UK, making London the largest office outside of San Francisco. But it would be hasty to read into this any nod towards those courting the company to move here. A couple of days before the event, the business said it plans to create just as many jobs in Dublin by 2027. Last year it also opened outposts in Paris and Munich.
I had gone in expecting to see some hard lobbying. A few years ago, I often went to events with this kind of set-up that were organised by crypto companies. They would book out plush hotel function rooms and secure ministerial keynotes and wear suits, all to demonstrate that they had adults in the room.
The purpose of those was to woo regulators and lawmakers, putting the hard sell on them in terms of what blockchain could do for the country, and showing willingness – even eagerness – to be regulated.
Yet Anthropic played things a lot softer. There was no laundry list for ministers. Instead, government was spoken of as a partner.
(I’m writing in a slightly roundabout, passive construction here because the event was conducted under Chatham House Rules. That means I can tell you a bit about what was said, but not who said it or who else was there.)
Anthropic already has partnerships with the UK Government. This originated in its work with the AI Security Institute (AISI) a memorandum signed last year, in which Anthropic agreed to work on citizen-facing tools and to explore other areas such as AI for science. Earlier this year, the company was selected to provide chat services within the new GOV.UK app.
But between the 101 explanation of agentic AI, the run-down of current safety thinking, and the live demos of products, one got the sense that the company would like a greater presence across more departments and agencies. After all, the public sector is ramping up its spending on AI contracts.
One of the demo stands, which showed off the Claude plug-in for Excel, was even named ‘In Whitehall’. I thought this was especially interesting because if you look at why some AI tools have failed to catch on in the civil service, it often comes down to a lack of integration with existing systems. Anthropic may be able to compete with CoPilot as the assistant of choice, but there is little chance of displacing Microsoft on the base layer, so it’s best to pitch on that level.
Other stands demonstrated multi-agent systems, how Claude might be used in life sciences labs, and how it is being rolled out in schools. For the latter, Anthropic has a partnership with Denver-based MagicSchool AI. Only a handful of schools in the UK are using its products, but the company uses Claude to serve several million users in the US.
And then there was the chance to use a Poetry Camera, a viral Kickstarter project that uses Claude to compose poetry based on images. This had nothing to do with policy, but it was fun to try. Even if I did think the sonnet it made me about the ceiling of Banqueting House was a bit naff.
All in all, it was an interesting combination of a proud display with a softly-softly approach. A contradiction in some ways, but one that feels suited to the brand. Perhaps – to break my rule of never using this particular adverb since it’s been overused by LLMs – it was a sign they are quietly confident.
Teatime scroll Each week I share links to writings, events, tweets and other conversation-starters. If you have something you think should be in here, feel free to reply to this email or DM me.
🥣 Congratulations to Granola on becoming a Unicorn after its latest funding round. I interviewed Chris Pedregal back in late 2024, when we talked about the (then mostly theoretical) idea of an AI workspace, big companies versus startups, and the importance of writing for thinking.
🦄 Frontier Tower and the Unicorn Mafia are teaming up to host a two-week ‘Agents in the Wild’ build sprint, culminating in a demo night on 14th April. Full details and the application form are here.
💡 Thank you to Tom Bristow of The Morning Intelligence for his help getting me into the event featured in today’s edition. If you haven’t already, do sign up for his daily briefing newsletter on AI in the UK.
📴 Pilots are underway in the UK testing different approaches to social media restrictions for under-16s.
🧬 Edith Heard, leader of the Francis Crick Institute, says “everyone should be worried” about investment in basic science amid pressure to focus on applied research, in an interview with Frances Jones for Research Fortnight.
🍻 If you’ve been following the drama over alleged AI usage in Matt Goodwin’s new book, you can watch him debate the matter with Andy Twelves tonight on GB News. There’s even a viewing party happening in Westminster.
🏛️ I will be in DC for the next few weeks, so the newsletter may come out at a slightly different time and I’ll hopefully be bringing you a few relevant tidbits from the other side of the pond. If there’s someone you think I should meet out there, get in touch.
Thanks for this - i want to keep up with ai in the uk :)