Excited about our new paper: AI Agent Traps AI agents inherit every vulnerability of the LLMs they're built on - but their autonomy, persistence, and access to tools create an entirely new attack surface: the information environmental itself. The web pages, emails, APIs, and databases agents interact with can all be weaponised against them. We introduce a taxonomy of six classes of adversarial threats - from prompt injections hidden in web pages to systemic attacks on multi-agent networks. I’m outlining the six categories of traps in the thread bellow

Some news: I recently started at @GoogleDeepMind w/ SAMBA (sociotechnical analysis of model behavior and alignment) w/ @kldivergence, @wsisaac, @charvvvv_, @kweku_gdm & many others! We're working on eval science & econ impacts. E.g., simulating real-world job-tasks to understand model capabilities & labor market impacts; countering benchmark hacking & data leakage, etc. Thrilled to help push this exciting research forward w an amazing team!

Looks like good reading for my train ride tomorrow.

Our report "From AGI to ASI" is out!

Well this goes instantly on the to-read list.

New Report: From AGI to ASI. arxiv.org/abs/2606.12683 The shift from human-level AI to superintelligence will be continuous accelerating breakthroughs across science & tech. We must prepare for a series of transformative societal changes. #AGI #ASI

AGI is sometimes framed as a single end point; more likely are successive "waves" of technological transformation... 🌊🌊🌊🌊 For more on this topic—and the possibility of Artificial Superintelligence—check out this @GoogleDeepMind report, led by Tim Genewein!

How do we go from AGI to Superintelligence? New report discusses four potential pathways: scaling, AI paradigm shifts, recursive improvement, and ASI emerging from large-scale multi- agent collectives. Importantly, it also looks at possible frictions and bottlenecks along these pathways. Instant classic! arxiv.org/abs/2606.12683

Thanks Socrates! 2.5k years ago you argued against that the hot new technology of the day (writing) would ruin everyone’s memory. Glad to see you’re keeping the tradition alive!!

apply!

Applications are due by 23:59 AoE on 9 August 2026. Find out more and submit your proposal here: cooperativeai.com/grants/mul….

Funding opportunity in multi-agent safety! I'm especially interested in emergent communication, but many important problems are represented:

How do we ensure safety in a world with millions of interacting agents, built and deployed by many different actors? Learn more about our new research fund in partnership with @GoogleDeepMind @Googleorg @coop_ai @ARIA_research: schmidtsciences.org/multi-ag…

Excited that this is out: $10M research funding towards multi-agent AI safety and security. A lot of work to be done -- apply by 9 August!

Academic and independent researchers should apply by August 8, 2026. Let's build a safe foundation for a multi-agent future. Read more about the funding call and apply now: deepmind.google/blog/investi…

With @schmidtsciences, @coop_ai, @ARIA_research and @Googleorg, we're launching a funding call for multi-agent AI safety. As AI scales, we'll move from models in isolation to complex ecosystems of interacting agents. This creates new risks that we need to solve.

AI is rapidly evolving from individual models to ecosystems of interacting agents that will communicate, negotiate and make decisions together at scale. Ensuring these systems can be trusted to act on our behalf is one of the defining opportunities of the coming decade. That’s why ARIA is launching a joint funding call with Schmidt Sciences, Google DeepMind, the Cooperative AI Foundation, and Google.org in a new $10M research fund focused on multi-agent AI safety and security. By bringing together a community of independent researchers around the world, we aim to build the infrastructure for multi-agent AI systems to securely coordinate, negotiate, and verify with one another on our behalf. Read more about the call: aria.org.uk/opportunity-spac…

Together with @schmidtsciences, @GoogleDeepMind, @ARIA_research, and supported by @Googleorg, we’re announcing a new $10m fund to support the foundational scientific research needed to understand, evaluate, and control risks emerging from large-scale ecosystems of interacting AI agents, deployed by multiple actors.

Some powerful reflections from my colleague @bakkermichiel team, on the role of Europe and possible AI futures. Few people are better placed to understand these dynamics!

"What will happen to Europe if it keeps ignoring AI?" Three American labs each (!!) operate more AI compute than all of Europe combined. Today we're launching Europe 2031: a story of what might happen if that doesn't change.

$10MM research fund for multi-agent AI safety -- link to apply below!