Control Red Team - Research Engineer/Research Scientist
UK AI Security Institute · London, UK · London
CHECKED 19H AGO
Research instituteonsite
At a glance
Highlights
- Stress-test control monitors from frontier labs including GDM and Anthropic.
- Design experiments measuring whether controls catch capable models causing harm.
- Break monitors, sandboxes and escalation protocols to guide government action.
What they offer
- Frontier model access with serious compute and infrastructure support
- Early role with real ownership and chance to publish research
- Privileged insight into control measures across frontier developers
Ideal candidate
A curious, adversarial thinker excited by AI safety who enjoys breaking systems, running experiments, and debating tricky measurement questions with a small red team.
- visa sponsorship
- Open to applicants of any nationality or background
- hiring process
- Pre-employment checks including criminal record check and BPSS clearance, with preference for CTC eligibility.
Lab profile
About UK AI Security Institute
UK government institute studying advanced AI risks and turning research into protective action.
Known for
- Alignment red-teaming
- Control red-teaming
- Misuse red-teaming
- Cyber and autonomous systems security
Key facts
- Based in the UK
- UK government-backed
- World's largest dedicated advanced AI risk team