Programme Research Outputs

We conduct and support research aiming to reduce risks posed by advanced AI both directly through our Alignment Research team, and though our training programmes. Since 2023, LASR Labs and the AI Governance Taskforce have supported 43 projects, with 195 contributors. Our research papers have been accepted to leading academic conferences and workshops including NeurIPS and ICML.

Highlights

ICML 2026

A loss curvature account of fine-tuning fragility

Ivaylo Dimitrov*, Leo Karoubi*, Sunny Howard*, Dmitrii Krasheninnikov.

View paper

LASR

ICML 2026

The Model Organism Lottery: Model Organism Interpretability Strongly Depends on Training Methodology

Andrzej Szablewski*, Gabriel Konar-Steenberg*, Raffaello Fornasiere*, Nikita Menon*, Stefan Heimersheim.

View paper

LASR

ICML 2026

One Probe Won't Catch Them All: Towards Targeted Deception Detection

Vikram Natarajan*, Devina Jain*, Shivam Arora*, Satvik Golechha, Joseph Bloom. Paper accepted to ICML 2026.

View paper

LASR

New research

Understanding as an Explicit and Assessable Component of Frontier AI Safety Decision

Steve Barrett (Research Team Leader), Alexandra Chirilǎ, Mamoon Masud, David Meredith Hardy, Phillip Mulvana
Expert Partner: Professor Robin Bloomfield (City University, University of London)

View paper

AI Governance Taskforce

New research

Testing and Evaluation of Agentic AI Systems in Military Command and Control

Expert Partner: Heather Frase (Veraitech), Ulysse Richard (Research Team Leader), Sarah Cao, Di Cooke, Sebastian Kwon, Adrianna Tan

View paper

AI Governance Taskforce

New research

Designing Verification and Enforcement for International AI Governance: Lessons from Historical Treaties

Expert Partner: Jose Villalobos (FLI - Future of Life Institute), Tara Thwing (Research Team Leader), Alixandra Acker, Joel Christoph, Julian Theseira, Uma Kalkar

View paper

AI Governance Taskforce

New research

Mapping General-Purpose AI Governance in Twenty AI Middle-Power Jurisdictions

Expert Partner: Caio Machado (The Future Society), Josephine Schwab (Research Team Leader), Ferruccio Barazzutti, Sheryn Lee, Nathan Naidoo

View paper

AI Governance Taskforce

New research

Architecting Candor: Products Liability and AI Incident Knowledge Governance

Expert Partner: Sean McGregor (AIID- AI Incident Database), Michael Celone (Research Team Leader), Bekhzodkhon (Beck) Alikhanov, Mosi Secret, Eduardo Mignot, Noga Bregman

View paper

AI Governance Taskforce

New research

Capability-Based Planning for AI Crisis Preparedness

Expert Partner: Elizabeth Dearden-Williams (UK AISI), Isaak Mengesha (Research Team Leader), Charlie Collins, Juan Felipe Ceron Uribe, Victoria Ellis, Salvatore d'Ambrosio

View paper

AI Governance Taskforce

ICML TAIGR 2026

Designing escalation criteria for international AI incident response: criteria, triggers, and thresholds

Francesca Gomez (Research Team Leader), Josephine Schwab, Michael Harré, Lydia Preston, Matthew Ball

View paper

AI Governance Taskforce

ICML TAIGR 2026

A Pragmatic Classification Framework for AI Incident Monitoring

Isaak Mengesha (Research Team Leader), Branwen Owen, Charlie Collins, Tina Wong

View paper

AI Governance Taskforce

Best paper award at ICML TAIGR 2026

Lessons from External Review of DeepMind’s Scheming Inability Safety Case

Steve Barrett (Research Team Leader), Javier Campos, Sean Fillingham, Umair Siddique, James Walpole

View paper

AI Governance Taskforce

ICML TAIGR 2026

Time magazine article

AI Incident Monitoring through a Public Health Lens

Simon Mylius (Research Team Leader), Giovanna Jaramillo-Gutierrez (Research Team Leader), Cyril Chuun, Sophia Abraham, Taiye (Chan) Chen, Sayash Raaj

View paper

AI Governance Taskforce

NeurIPS 2025

CoT Red-Handed: Stress Testing Chain-of-Thought Monitoring

Benjamin Arnav, Pablo Bernabeu-Pérez, Nathan Helm-Burger, Tim Kostolansky, Hannes Whittingham, Mary Phuong

View paper

LASR

NeurIPS 2024

Hidden in Plain Text: Emergence and Mitigation of Steganographic Collusion in LLMs

Yohan Mathew*, Ollie Matthews*, Robert McCarthy*, Joan Velja*, Christian Schroeder de Witt, Dylan Cope, Nandi Schoots

View paper

LASR

NeurIPS 2024

Characterizing stable regions in the residual stream of LLMs

Jett Janiak, Jacek Karwowski*, Chatrik Singh Mangat*, Nora Petrova, Giorgi Giglemiani, Stefan Heimersheim

View paper

LASR

Best paper award at ICML 2025; NeurIPS 2025

Detecting High-Stakes Interactions with Activation Probes

Alex McKenzie, Urja Pawar, Phil Blandfort, William Bankes, David Krueger, Ekdeep Singh Lubana, Dmitrii Krasheninnikov

View paper

LASR

NeurIPS 2024

Evaluating Synthetic Activations composed of SAE Latents in GPT-2

Giorgi Giglemiani*, Nora Petrova*, Chatrik Singh Mangat*, Jett Janiak, Stefan Heimersheim

View paper

LASR

Stress-Testing Alignment Midtraining

Sid Baines, Jonathan Bostock, Maria Angelica Martinez, Andrew Draganov, David Africa, Daniel Tan

View paper

Alignment

scroll for more

Full research outputs

AI Governance Taskforce

Understanding as an Explicit and Assessable Component of Frontier AI Safety Decision

Steve Barrett (Research Team Leader), Alexandra Chirilǎ, Mamoon Masud, David Meredith Hardy, Phillip Mulvana
Expert Partner: Professor Robin Bloomfield (City University, University of London)

AI Governance Taskforce

Understanding as an Explicit and Assessable Component of Frontier AI Safety Decision

Steve Barrett (Research Team Leader), Alexandra Chirilǎ, Mamoon Masud, David Meredith Hardy, Phillip Mulvana
Expert Partner: Professor Robin Bloomfield (City University, University of London)

AI Governance Taskforce

Understanding as an Explicit and Assessable Component of Frontier AI Safety Decision

Steve Barrett (Research Team Leader), Alexandra Chirilǎ, Mamoon Masud, David Meredith Hardy, Phillip Mulvana
Expert Partner: Professor Robin Bloomfield (City University, University of London)

AI Governance Taskforce

Testing and Evaluation of Agentic AI Systems in Military Command and Control

Expert Partner: Heather Frase (Veraitech), Ulysse Richard (Research Team Leader), Sarah Cao, Di Cooke, Sebastian Kwon, Adrianna Tan

AI Governance Taskforce

Testing and Evaluation of Agentic AI Systems in Military Command and Control

Expert Partner: Heather Frase (Veraitech), Ulysse Richard (Research Team Leader), Sarah Cao, Di Cooke, Sebastian Kwon, Adrianna Tan

AI Governance Taskforce

Testing and Evaluation of Agentic AI Systems in Military Command and Control

Expert Partner: Heather Frase (Veraitech), Ulysse Richard (Research Team Leader), Sarah Cao, Di Cooke, Sebastian Kwon, Adrianna Tan

AI Governance Taskforce

Designing Verification and Enforcement for International AI Governance: Lessons from Historical Treaties

Expert Partner: Jose Villalobos (FLI - Future of Life Institute), Tara Thwing (Research Team Leader), Alixandra Acker, Joel Christoph, Julian Theseira, Uma Kalkar

AI Governance Taskforce

Designing Verification and Enforcement for International AI Governance: Lessons from Historical Treaties

Expert Partner: Jose Villalobos (FLI - Future of Life Institute), Tara Thwing (Research Team Leader), Alixandra Acker, Joel Christoph, Julian Theseira, Uma Kalkar

AI Governance Taskforce

Designing Verification and Enforcement for International AI Governance: Lessons from Historical Treaties

Expert Partner: Jose Villalobos (FLI - Future of Life Institute), Tara Thwing (Research Team Leader), Alixandra Acker, Joel Christoph, Julian Theseira, Uma Kalkar

AI Governance Taskforce

Architecting Candor: Products Liability and AI Incident Knowledge Governance

Expert Partner: Sean McGregor (AIID- AI Incident Database), Michael Celone (Research Team Leader), Bekhzodkhon (Beck) Alikhanov, Mosi Secret, Eduardo Mignot, Noga Bregman

AI Governance Taskforce

Architecting Candor: Products Liability and AI Incident Knowledge Governance

Expert Partner: Sean McGregor (AIID- AI Incident Database), Michael Celone (Research Team Leader), Bekhzodkhon (Beck) Alikhanov, Mosi Secret, Eduardo Mignot, Noga Bregman

AI Governance Taskforce

Architecting Candor: Products Liability and AI Incident Knowledge Governance

Expert Partner: Sean McGregor (AIID- AI Incident Database), Michael Celone (Research Team Leader), Bekhzodkhon (Beck) Alikhanov, Mosi Secret, Eduardo Mignot, Noga Bregman

AI Governance Taskforce

Mapping General-Purpose AI Governance in Twenty AI Middle-Power Jurisdictions

Expert Partner: Caio Machado (The Future Society), Josephine Schwab (Research Team Leader), Ferruccio Barazzutti, Sheryn Lee, Nathan Naidoo

AI Governance Taskforce

Mapping General-Purpose AI Governance in Twenty AI Middle-Power Jurisdictions

Expert Partner: Caio Machado (The Future Society), Josephine Schwab (Research Team Leader), Ferruccio Barazzutti, Sheryn Lee, Nathan Naidoo

AI Governance Taskforce

Mapping General-Purpose AI Governance in Twenty AI Middle-Power Jurisdictions

Expert Partner: Caio Machado (The Future Society), Josephine Schwab (Research Team Leader), Ferruccio Barazzutti, Sheryn Lee, Nathan Naidoo

AI Governance Taskforce

Capability-Based Planning for AI Crisis Preparedness

Expert Partner: Elizabeth Dearden-Williams (UK AISI), Isaak Mengesha (Research Team Leader), Charlie Collins, Juan Felipe Ceron Uribe, Victoria Ellis, Salvatore d'Ambrosio

AI Governance Taskforce

Capability-Based Planning for AI Crisis Preparedness

Expert Partner: Elizabeth Dearden-Williams (UK AISI), Isaak Mengesha (Research Team Leader), Charlie Collins, Juan Felipe Ceron Uribe, Victoria Ellis, Salvatore d'Ambrosio

AI Governance Taskforce

Capability-Based Planning for AI Crisis Preparedness

Expert Partner: Elizabeth Dearden-Williams (UK AISI), Isaak Mengesha (Research Team Leader), Charlie Collins, Juan Felipe Ceron Uribe, Victoria Ellis, Salvatore d'Ambrosio

LASR

The Model Organism Lottery: Model Organism Interpretability Strongly Depends on Training Methodology

Andrzej Szablewski*, Gabriel Konar-Steenberg*, Raffaello Fornasiere*, Nikita Menon*, Stefan Heimersheim.

LASR

The Model Organism Lottery: Model Organism Interpretability Strongly Depends on Training Methodology

Andrzej Szablewski*, Gabriel Konar-Steenberg*, Raffaello Fornasiere*, Nikita Menon*, Stefan Heimersheim.

LASR

The Model Organism Lottery: Model Organism Interpretability Strongly Depends on Training Methodology

Andrzej Szablewski*, Gabriel Konar-Steenberg*, Raffaello Fornasiere*, Nikita Menon*, Stefan Heimersheim.

LASR

One Probe Won't Catch Them All: Towards Targeted Deception Detection

Vikram Natarajan*, Devina Jain*, Shivam Arora*, Satvik Golechha, Joseph Bloom. Paper accepted to ICML 2026.

LASR

One Probe Won't Catch Them All: Towards Targeted Deception Detection

Vikram Natarajan*, Devina Jain*, Shivam Arora*, Satvik Golechha, Joseph Bloom. Paper accepted to ICML 2026.

LASR

One Probe Won't Catch Them All: Towards Targeted Deception Detection

Vikram Natarajan*, Devina Jain*, Shivam Arora*, Satvik Golechha, Joseph Bloom. Paper accepted to ICML 2026.

AI Governance Taskforce

Strengthening the International Network for Advanced AI Measurement, Evaluation and Science to Support Enforceable Global AI Red Lines

Tara Thwing (Research Team Leader), David Gringras, Ulysse Richard, Matthew Tang

AI Governance Taskforce

Strengthening the International Network for Advanced AI Measurement, Evaluation and Science to Support Enforceable Global AI Red Lines

Tara Thwing (Research Team Leader), David Gringras, Ulysse Richard, Matthew Tang

AI Governance Taskforce

Strengthening the International Network for Advanced AI Measurement, Evaluation and Science to Support Enforceable Global AI Red Lines

Tara Thwing (Research Team Leader), David Gringras, Ulysse Richard, Matthew Tang

LASR

EvalDetectBench: A Benchmark for Measuring Evaluation Awareness in Frontier Language Models

Xinning Li*, Kemunto Ochwang'i*, Aryasomayajula Ram Bharadwaj*, Alexandra Souly, Robert Kirk.

LASR

EvalDetectBench: A Benchmark for Measuring Evaluation Awareness in Frontier Language Models

Xinning Li*, Kemunto Ochwang'i*, Aryasomayajula Ram Bharadwaj*, Alexandra Souly, Robert Kirk.

LASR

EvalDetectBench: A Benchmark for Measuring Evaluation Awareness in Frontier Language Models

Xinning Li*, Kemunto Ochwang'i*, Aryasomayajula Ram Bharadwaj*, Alexandra Souly, Robert Kirk.

LASR

Phantom Transfer: Data-level Defences are Insufficient Against Data Poisoning

Andrew Draganov*, Anandmayi Bhongade*, Tolga Dur*, Mary Phuong.

LASR

Phantom Transfer: Data-level Defences are Insufficient Against Data Poisoning

Andrew Draganov*, Anandmayi Bhongade*, Tolga Dur*, Mary Phuong.

LASR

Phantom Transfer: Data-level Defences are Insufficient Against Data Poisoning

Andrew Draganov*, Anandmayi Bhongade*, Tolga Dur*, Mary Phuong.

AI Governance Taskforce

Signals in the Noise: Open Source Intelligence (OSINT) for AI Loss of Control Detection

Sarah Bollinger (Research Team Leader), Nada Aboserie, Ian Lee, Taysir Mathlouthi, Amanda Coakley

AI Governance Taskforce

Signals in the Noise: Open Source Intelligence (OSINT) for AI Loss of Control Detection

Sarah Bollinger (Research Team Leader), Nada Aboserie, Ian Lee, Taysir Mathlouthi, Amanda Coakley

AI Governance Taskforce

Signals in the Noise: Open Source Intelligence (OSINT) for AI Loss of Control Detection

Sarah Bollinger (Research Team Leader), Nada Aboserie, Ian Lee, Taysir Mathlouthi, Amanda Coakley

LASR

A loss curvature account of fine-tuning fragility

Ivaylo Dimitrov*, Leo Karoubi*, Sunny Howard*, Dmitrii Krasheninnikov.

LASR

A loss curvature account of fine-tuning fragility

Ivaylo Dimitrov*, Leo Karoubi*, Sunny Howard*, Dmitrii Krasheninnikov.

LASR

A loss curvature account of fine-tuning fragility

Ivaylo Dimitrov*, Leo Karoubi*, Sunny Howard*, Dmitrii Krasheninnikov.

AI Governance Taskforce

Designing escalation criteria for international AI incident response: criteria, triggers, and thresholds

Francesca Gomez (Research Team Leader), Josephine Schwab, Michael Harré, Lydia Preston, Matthew Ball

AI Governance Taskforce

Designing escalation criteria for international AI incident response: criteria, triggers, and thresholds

Francesca Gomez (Research Team Leader), Josephine Schwab, Michael Harré, Lydia Preston, Matthew Ball

AI Governance Taskforce

Designing escalation criteria for international AI incident response: criteria, triggers, and thresholds

Francesca Gomez (Research Team Leader), Josephine Schwab, Michael Harré, Lydia Preston, Matthew Ball

AI Governance Taskforce

A Pragmatic Classification Framework for AI Incident Monitoring

Isaak Mengesha (Research Team Leader), Branwen Owen, Charlie Collins, Tina Wong

AI Governance Taskforce

A Pragmatic Classification Framework for AI Incident Monitoring

Isaak Mengesha (Research Team Leader), Branwen Owen, Charlie Collins, Tina Wong

AI Governance Taskforce

A Pragmatic Classification Framework for AI Incident Monitoring

Isaak Mengesha (Research Team Leader), Branwen Owen, Charlie Collins, Tina Wong