Tech and better futures. Lead Researcher (Digital Policy) @Demos | Ex @collect_intel, @turinginst, @_GlobalAssembly
London
Joined November 2012
- Tweets472
- Following1.3K
- Followers505
- Likes1.3K
Flynn Devine retweeted
AISI IS HIRING!
It was less than a month ago that I became Director. I joined with the belief that AISI is a world-leading organisation, and living proof that government can build things that work.
One month in and I'm even more bullish about the vital role @AISecurityInst plays in frontier AI security. I'm astounded daily by the talent, focus and commitment of the brilliant team I get to work with.
There has never been a more urgent time to work on frontier AI security - independently, in the public interest. We have a lot of urgent work to do, our team is growing fast, and we need more brilliant people to fill those roles.
Our Red Team uses adversarial machine learning techniques to find failures in alignment measures, control monitors, and misuse guardrails. They are massively scaling up all parts of the team:
Alignment: job-boards.eu.greenhouse.io/…
Control: job-boards.eu.greenhouse.io/…
Misuse: job-boards.eu.greenhouse.io/…
AI capabilities in cybersecurity and autonomy are advancing faster than ever. Our Cyber & Autonomous Systems team assesses what frontier and open-weight models can really do - across cyber, autonomy and AI R&D. We're hiring a Cyber Security Engineer and Software Engineer to build the evaluations behind that work:
CSE: job-boards.eu.greenhouse.io/…
SWE: job-boards.eu.greenhouse.io/…
Our Human Influence team studies how AI can shift human decisions and behaviour, using everything from RCTs to multi-agent studies. We're hiring an Engineering Lead to help us grow the ambition and pace of that research:
job-boards.eu.greenhouse.io/…
We're also hiring exceptional Software Engineers at all seniority levels to join our Core Technology team, which works closely with researchers to build performant tools and infrastructure that enable and accelerate AISI's world-leading AI safety research:
job-boards.eu.greenhouse.io/…
Flynn Devine retweeted
Ahead of big socioeconomic shifts, you often see the emergence of small prophetic sects. Most vanish into obscurity or purity spirals, but sometimes their members escape their narrow confines and built durable institutions.
@lawhsw and I explore how this needs to happen for AI.
"Reading METR graphs does not automatically bring with it an understanding of how specific domains will respond to agent ecologies."
@chalmermagne and @lawhsw on why AI institution-builders need to think bigger about the future they're building for.
blog.cosmos-institute.org/p/…
Incredible raise from @DanAkarca, @achterbrain and the @CallosumAI team! Exciting for the UK AI ecosystem.
Today we're announcing our $100m seed round to change how humanity computes 🚀
AI must be far cheaper, faster and more performant than it is today. Co-evolving heterogeneous compute and intelligence is how we do it.
With our round, today we are announcing global partnerships, new technical advances, and the launch of Tailored Inference to optimise tasks, models and hardware simultaneously.
A future of heterogeneous intelligence, built on principles we see in nature, is coming fast. And we're just getting started!
My new paper in @Data_and_Policy with Nathan Davies argues for a political economy lens on how we use AI to engage citizens in politics.
Technical affordances are important, but lacking a broader analysis our thinking is limited.
Commentary published in Data & Policy journal!
AI and citizen participation: a political economy lens
👉cup.org/4gaeaLf
By Nathan Davies and @FlynnDevine
#AI #citizenparticipation #governance
Flynn Devine retweeted
Commentary published in Data & Policy journal!
AI and citizen participation: a political economy lens
👉cup.org/4gaeaLf
By Nathan Davies and @FlynnDevine
#AI #citizenparticipation #governance
Flynn Devine retweeted
Published today by @Demos, new evidence of 'Retrieval Augmented Generation (RAG) poisoning’. The deliberate manipulation of online information so that AI models amplify false or misleading narratives.
Specifically, by using GEO (Generative Engine Optimisation), which is used to make content online more likely to come up in AI searches.
This is an important type of AI epistemic risk. GEO is an extension of / mimicking SEO (search engine optimisation), which has been used for decades.
"The number of GEO-providers offering these techniques, including PR and SEO companies, suggests there is a relatively low barrier to offering such services. Given these services are now so readily available, suggests a broad range of actors could access the skills to influence LLM-powered AI services, including state actors."
The report identifies this as a growing threat.
They use a case study of foreign information manipulation and interference (FIMI). Their research demonstrates how these techniques are *already* influencing the outputs of leading AI services.
The report argues that protecting the UK’s information environment now requires securing AI systems as well as the wider digital ecosystem.
The report sets out practical recommendations, and explains the specific things malicious actors are doing to distort the information ecosystem.
@Demos we tested a series of AI chat bots on elections during an election period and the results are concerning.
34.1% of responses included errors including inventing candidates, giving wrong election dates, and making up scandals.
Full report linked below.
Really interesting resource released by @windfalltrust mapping policy responses to economic disruption from AI that may be coming down the tracks.
Important not just as record of good ideas but as a means to connect ideas to inform a whole of society approach.
Link below.
This is incredibly cool. Always impressed by the work @saffronhuang and the team are doing.
Interesting manifestation of the AI abundance dreams coming through with a wish for more work freedom. Wonder how this trades off against total future of work anxiety on grand scale.
Flynn Devine retweeted
SOOO excited to share this research. It has been many months in the making, and it’s the largest and certainly the most multilingual qualitative study that we think has ever been run!?
We invited Claude users to share how they use AI, what they dream it could make possible, and what they fear it might do.
Nearly 81,000 people responded in one week—the largest qualitative study of its kind.
Read more: anthropic.com/features/81k-i…
Opportunity like no other to shape the future of UK R&D through moonshot bets.
The search for ARIA’s next cohort of Programme Directors has begun 🚀
As an ARIA PD, you will design and actively manage a ~£50M R&D programme within or around our opportunity spaces with the goal of unlocking scientific and technological breakthroughs that benefit everyone.
There is no one way to be an ARIA PD – our existing cohorts come from a range of backgrounds, including academia, entrepreneurship, invention and industry, and have launched programmes in areas ranging from synthetic plants to multi-agent coordination to brain surgery-free neurotechnologies.
Full applications will open in August 2026 for a May 2027 start date – over the coming months we’ll be running webinars and in-person events across the UK, Europe, the US and Asia where you’ll get the chance to engage directly with the ARIA team and learn more about the opportunity.
Find out more about what it means to be a PD and register your interest to be the first to receive updates on the recruitment process: link.aria.org.uk/pdc3-x
Facilitated @windfalltrust workshop on AI-caused economic collapse month ago and been thinking about it non stop.
A lot of hype right now with viral essays abound but no doubt we should at least take these scenarios into serious consideration.
Coordination problem in action.
“We didn't really feel, with the rapid advance of AI, that it made sense for us to make unilateral commitments … if competitors are blazing ahead.”
Exclusive: Anthropic is dropping the central pledge of its flagship safety policy, company officials tell TIME.
time.com/7380854/exclusive-a…
Flynn Devine retweeted
Raspberry Pi stock up 63% since this speech btw 🤷♂️
Last week, I set out this Government’s vision on AI for our startup ecosystem: Machines That Serve Us.
It’s now up at 👉🏽 brit.dev/
The UK is a global top 3 $1tn tech ecosystem, focused now on moving faster still:
✅ A heritage of tech extending human agency
✅ Spreading AI opportunity across all our communities
✅ No more tall poppy syndrome, just a distinctly British AI trajectory
The Silicon Valley essay aesthetic making its way into UK politics - not totally against it.
Great to see latest from @KanishkaNarayan outlining UK strengths and ambition for AI and tech.
Last week, I set out this Government’s vision on AI for our startup ecosystem: Machines That Serve Us.
It’s now up at 👉🏽 brit.dev/
The UK is a global top 3 $1tn tech ecosystem, focused now on moving faster still:
✅ A heritage of tech extending human agency
✅ Spreading AI opportunity across all our communities
✅ No more tall poppy syndrome, just a distinctly British AI trajectory
@Demos is hiring a new Chair for our board.
Looking for a senior, ambitious, high agency person who can bring real insight to our work on strengthening democracy for the 21st century.
If you know someone who could be great, tag them below. More info in comments.
Flynn Devine retweeted
In 2025, "automating AI research" went from abstract idea to common practice inside leading AI companies.
Out today: ✨new report✨ from a workshop we ran with some of the smartest people on this topic.
🧵:
cset.georgetown.edu/publicat…
As someone who lives in (and loves) London, I’ve noticed the rise online and IRL of folks saying it’s dangerous and collapsing.
The city has its problems, but does feel like this shift isn’t a coincidence…
Interesting take from @Hothi and @trustforlondon in comments.
