A few quick takes: 1) Evals like this are windows into the future, which makes them very valuable. The most concerning actions (spear phishing, trying to inject malicious code) were done by Mythos 5, which is unreleased. 1/5 www.aisi.gov.uk/blog/inciden...

We are entering a new world. The implications and consequences will require all our best thinking to handle, but no scientist (myself included) could help but feel child-like wonder at the possibilities. openai.com/index/ten-ad...

Ten advances in mathematics and theoretical computer science

OpenAI shares new results on long-standing open problems in mathematics and theoretical computer science, including advances in geometry, cryptography, and complexity.

openai.com

I was grateful to get an early read of this excellent new work by FAR AI on the relative ease/difficulty of jailbreaking frontier models. Really eye-opening the difference between the robustness of OpenAI/Anthropic's models vs Gemini and Grok. 1/2 finance.yahoo.com/technology/a...

FAR.AI Launches AI Security Leaderboard Revealing Hundredfold Gap in Frontier AI Model Safeguards

AI security research nonprofit FAR.AI today launched its AI Security Leaderboard at leaderboard.far.ai, a robust independent public ranking of how well the safeguards on frontier AI models withstand t...

finance.yahoo.com

I was pleased to be interviewed for this excellent article by Robert Hart. Policymakers and the public should be taking the OpenAI loss of control incident, and the underlying trend it is part of, seriously. www.theverge.com/ai-artificia...

We’re running out of reasons to ignore AI safety

In the aftermath of OpenAI’s attack on Hugging Face, experts say it’s time for everyone to take security far more seriously.

theverge.com

'As models keep getting more capable and harder to contain, the field would benefit from more outside visibility into what happens inside AI labs before something goes wrong, Ó hÉigeartaigh said, rather than learning about it only after.' www.msn.com/en-us/techno...

Incredible research by Antonia Juelich: Boko Haram's use of frontier AI: www.nytimes.com/2026/07/10/u...

How Terrorist Groups Are Using A.I. to Gain an Edge in Battle

A.I. chatbots are not just a propaganda tool for violent extremists but are aiding in bomb construction and attack planning, new research finds.

nytimes.com

A thoughtful and generous response by Matthew Willis to Giulio Corsi and my commentary on their paper in Philosophy and Technology. I'm particularly grateful to the journal for supporting this kind of academic dialogue; I think it's very helpful. link.springer.com/article/10.1...

Caring for Disagreement: Response to ÓhÉigeartaigh and Corsi - Philosophy & Technology

This is a response to “Bridging Divides and Building Resilience: a Reply to Bridging the Divide: Using LLMs to Reshape Difficult Disagreements” by ÓhÉigeartaigh and Corsi. While I agree that LLM-based...

link.springer.com

Thoughtful essay from Andreas Kirsch about corporate governance, surveillance and military uses of AI, and the Google/Pentagon contract. Worth a read. utaw.tech/news/trust-i...

UTAW: Trust is not Governance

DeepMind has bet that a strong safety culture and good leadership built on trust are sufficient to withstand outside pressure. The bet has failed.

utaw.tech

It was a genuine pleasure to provide feedback on these excellent and detailed AI scenarios now published by GO Science (with input from DSIT, AISI and external experts) - I was particularly impressed by how seriously they took their premises. www.gov.uk/government/p...

AI Scenarios 2030: Helping policymakers plan for the future of AI

gov.uk

Quite exciting that our international team at CFI provides 5% (and in fact, fully 10% if you count associate fellows and former CFI staff) of the entire EU AI Act Scientific Panel. Lovely of the School of Arts and Humanities to highlight it. www.csah.cam.ac.uk/news/cambrid...

I have bad news for you about the foundations of the rationalist ideology that is very influential in Silicon Valley.

This chapter is now out in a superb new volume 'The Artificial Intelligence Revolution', eds Max Rangeley & Lord Nicholas Fairfax. link.springer.com/book/10.1007... Contributions from leading thinkers including Yoshua Bengio, Andrew Yao, Barbara Oakley, Yi Zeng, Marc Warner, Terry Sejnowski, 1/2

Sean O hEigeartaigh@sean-o-h.bsky.social · last yr.

New working paper (pre-review), maybe my most important in recent years. I examine the evidence for the US-China race to AGI and decisive strategic advantage, & analyse the impact this narrative is having on our prospects for cooperation on safety. 1/5 papers.ssrn.com/sol3/papers....

Honoured to be appointed to the Scientific Panel for the EU AI Act. It's a really top-class group; I can think of few groups I've been so excited to work as part of. It's going to be really important to get this right. digital-strategy.ec.europa.eu/en/policies/...

www.wired.com/story/google...

Google DeepMind Workers Vote to Unionize Over Military AI Deals

UK staff of Google’s AI research lab hope to block the use of the company’s artificial intelligence models in military settings.

wired.com

Delighted to be a coauthor on this paper, which explores an approach for evaluating AI systems capable of continual learning. cl-eval.github.io

Thanks everyone who came along! I think it'll be online as a recorded version shortly, possibly on the same link. Really appreciated some excellent questions (plus the opportunity to read up in advance on developments in mirror bacteria, asteroid monitoring etc, alongside my usual fare of AI risk).

Sean O hEigeartaigh@sean-o-h.bsky.social · 3mo ago

Reminder that this is coming up this evening, 6-630pm UK! Drop along if you feel like it, open to all. cassyni.com/events/AKmfE...

Reminder that this is coming up this evening, 6-630pm UK! Drop along if you feel like it, open to all. cassyni.com/events/AKmfE...

I appreciated having the chance to speaking to Beatrice Nolan for this excellent piece on the Google/Pentagon contract. Grateful for journalists continuing to push for this to get the attention and scrutiny it deserved. fortune.com/2026/05/04/g...

Google’s AI deal with the Pentagon has sparked employee backlash. But don't expect a repeat of Project Maven | Fortune

Employee leverage over the tech giant's actions appears to have waned in the years since Project Maven.

fortune.com

Delight to be presenting my paper Extinction of the human species: What could cause it and how likely is it to occur? at Real World Challenges Seminar Series on 06 May 6pm UK. Anyone can join, link below. cassyni.com/events/AKmfE...

Pleased to have a commentary out today with Giulio Corsi in Philosophy and Technology. It's a response to a paper I'd recommend, Bridging the Divide: Using LLMs to Reshape Difficult Disagreements, 1/6

We’re proud to announce the winner of the 2025 Paper of the Year for #CambridgePrisms: Extinction🎉 Could humanity really face extinction? Read more #OA: https://cup.org/41e1jAo

🚨Hot off the press: A new study in Nature co-authored by CFI researchers introduces a methodology that can predict whether AI models will succeed or fail on unfamiliar tasks: www.lcfi.ac.uk/news-events/...

New methodology can predict whether AI will succeed on new tasks - LCFI

A new study led by the University of Cambridge just published in the Nature journal introduces a methodology that can predict how artificial intelligence (AI) models will perform on unfamiliar tasks, ...

lcfi.ac.uk

Pleased to share that my paper "Extinction of the human species: what could cause it and how likely is it to occur?" was awarded Cambridge Prisms: Extinctions (the journal)'s Paper of the Year award for 2025! www.cambridge.org/core/journal...

This is one of the new CFI developments I'm most excited about. Thrilled to have someone of Lucius' calibre here, and thrilled for us to be giving serious thought to this sort of question. www.lcfi.ac.uk/news-events/...

Cambridge Digital Minds launches at CFI - LCFI

New initiative - Cambridge Digital Minds - has launched at the Leverhulme Centre for the Future of Intelligence (CFI) to explore the societal, ethical, and governance implications of digital minds. As...

lcfi.ac.uk

A new pre-print "A Labour of Harm: Artificial Intelligence and Biological Weapons Acquisition" discusses how focusing on technological capability risks overlooks the human processes through which biological weapons are imagined Read here⬇️ papers.ssrn.com/sol3/papers....

Nice article on the Anthropic/pentagon saga, good work Harry and Billy time.com/article/2026...

The Most Disruptive Company in the World

How Anthropic's battle with the Pentagon became a fight for its future

time.com

'm getting this made up in glossy as a poster. It'll be on my wall next to my black-and-white photo mockup of Big Pete H benching, with the 'defective altruism' quote beneath it. Decades from now, I'll frame it and hang it in my loo l1/2

issuing correction on a previous post of mine, regarding the group effective altruism. you perhaps occasionally, under certain circumstances, "gotta hand it to them"

SE Gyges@segyges.bsky.social · 5mo ago

this is an attempt by the Secretary of War to bankrupt Anthropic. good luck to everyone working there

Excellent statement from the Center for Democracy and Technology’s @alexreevegivens.bsky.social, who notes that concern about today’s action is wide ranging and bipartisan www.linkedin.com/posts/alexan...

The Administration's move to cut all federal government use of Anthropic and designate it a Supply Chain Risk sets a dangerous precedent. There's a reason that national security leaders and… |...

The Administration's move to cut all federal government use of Anthropic and designate it a Supply Chain Risk sets a dangerous precedent. There's a reason that national security leaders and libertaria...

linkedin.com

RT from Jonathan Birch on twitter. "Say what you like about Effective Altruism, Pete Hegseth is not ranting about Kantians or virtue theorists trying to thwart his mass surveillance/autonomous weapons programs."