AI safety & policy
Full-time
Salary
£375,000 – £640,000 per year
Routes to impact
Direct high impact on an important cause
Description
Build and operate real-time safety classifiers that detect and block unsafe AI agent actions across training, evaluations, and internal deployments.
- Design, train, and evaluate classifiers, probes, and prompted models for rare or adversarial unsafe behaviour
- Deploy and tune classifiers for reliability, latency, cost, accuracy, and failure handling in critical systems
- Build datasets, labelling pipelines, evaluations, and harm taxonomies to measure and improve detection
- Partner with red teaming and other teams to stress-test classifiers and quickly resolve false positives
- Visa sponsorship available; hybrid role with at least 25% office attendance
This text was generated by AI. If you notice any inconsistencies, please let us know using this form.



