3 of 3 White-Box Testing Jobs in London

work of frontier AI labs, government agencies, and national security institutes.Our research spans fundamental research using black-box and white-box approaches to understand and steer AI systems, to building and advancing safety evaluations which better understand and quantify risks in AI. We care … designing and building AI evaluations or benchmarks, alongside a strong ability to reason about construct validity and mitigate confounds.Expertise in red-teaming, adversarial testing, jailbreaking, indirect prompt injection, and assessing the robustness of model safeguards.Strong foundational skills in experimental design, statistical analysis, and uncertainty quantification, including Bayesian methods.Deep ...

capable model attempting to cause harm is an empirical question that remains largely untested. The Control Red Team at the company is stress‐testing control monitors (including from GDM and Anthropic) – finding attacks that beat monitors and attempting to design experiments to measure real‐world risk. … that generate the evidence, writing arguments, and arguing them out with the rest of the team. We intend to publish this work. 2. Testing: Running evaluations of frontier labs' monitors, and reporting the implications. Turning our research into concrete assessments of real systems: threat modelling ...

capable model attempting to cause harm is an empirical question that remains largely untested. The Control Red Team at AISI is stress-testing control monitors (including from GDM and Anthropic) - finding attacks that beat monitors and attempting to design experiments to measure real-world risk. We're based … complex research projects involving substantial engineering. This includes black-box work (API-based evaluations and attacks) and ideally some white-box work (e.g. fine-tuning open-weight models). Strong software engineering and ML experience: writing clean, documented, reusable code for machine learning experiments ...