London AI Safety Research (LASR) Labs is an AI safety research programme run by @ArcadiaImpact focussed on reducing the risk of loss of control to advanced AI.
London
Joined April 2025
- Tweets22
- Following56
- Followers281
- Likes0
LASR Labs retweeted
Update! Our @LASRlabs paper, "Evaluating and Understanding Scheming Propensity in LLM Agents", has been accepted at NeurIPS 2026 Main Track for a poster presentation. See you all in Paris!
Was great working on this with @Jannoshh, @HopmanMia, @avramidou, and @davlindner!
An update on our @LASRlabs project on measuring scheming propensity!
How likely are frontier LLMs to actually scheme for self-preservation in realistic deployments?
Not very, as it turns out! Baseline scheming rates are ~0%, but tiny prompt changes can shift this to 70%+.
LASR Labs retweeted
I recently wrapped up my time as a @LASRlabs fellow during the Winter 2026 cohort (January to April) and did an extension from May to August, and it was one of the most rewarding experiences in my career. (1/4)
Jonathan's PhD is in DNA nanotech. Many of our strongest fellows arrive from fields adjacent to AI safety and machine learning. If you're technically strong, care about making AI go well, and are unsure you're qualified: apply.
Applications close in 6 days (20 Sept, 23:59 BST).
Irakli went from our most recent cohort to @AISecurityInst's Red Team within a few months. Here's what he had to say about it.
9 days left to apply! Applications close 20 September (23:59 BST).
Jannes first worked on scheming evals at LASR. He now works on scheming at Apollo Research.
Apply by 20 September (23:59 BST).
Apply by 20 Sept for the next LASR Labs cohort, running Jan–April in London.
13 weeks to write an AI safety paper on topics like scalable oversight, evaluation awareness and constitutions, with supervisors from @AISecurityInst, @GoogleDeepMind, @ApolloResearch, and more. (1/3)
Alumni have gone on to @AISecurityInst, @ApolloResearch, @GeodesResearch, @OpenAI's Alignment team and more, or continued researching with their supervisors.
Recent papers: three at NeurIPS 2025 main conference (one oral) and a best paper award at an ICML 2025 workshop. (2/3)
Details can be found here: lesswrong.com/posts/dzSAsL4f…
LASR Labs' website (and application) at: lasrlabs.org
🚨one week left to apply for the LASR Labs Winter 2026 Cohort!
Deadline is October 10th – don’t miss your chance to be part of something impactful.
Details can be found here: lesswrong.com/posts/zXugCu8A…
LASR Labs' website (and application) at: lasrlabs.org
Details can be found here: lesswrong.com/posts/zXugCu8A…
LASR Labs' website (and application) at: lasrlabs.org
Only 𝟱 𝗱𝗮𝘆𝘀 left to submit your application to the Summer 2025 AI safety research programme!
- 𝗽𝗮𝗶𝗱 𝗽𝗿𝗼𝗴𝗿𝗮𝗺𝗺𝗲 (£11,000 stipend)
- application only takes 20 minutes
- perfect for AI/ML people looking to break into AI safety, and have an impact
Good luck!
🚨Less than 2 weeks left to apply for the LASR Labs Summer 2025 Cohort!
⏳ Deadline: 𝐀𝐩𝐫𝐢𝐥 𝟐𝟔𝐭𝐡 – don’t miss your chance to be part of something impactful.
Apply by April 26th to LASR Labs AI Safety research program. This is a 3-month paid opportunity to write a paper in a small team with expert supervision. Past work published at workshops/conferences. Alumni now at Open AI & UKAISI. Details: lesswrong.com/posts/odZGKykS…