Description

You will lead a sub-team partnering with frontier AI companies to stress-test control measures for misaligned AI systems, using adversarial ML techniques to find failures and strengthen safety measures. Your work directly influences critical deployment and policy decisions within the UK government.

Requirements

• Expertise in adversarial ML and AI safety
• Experience in red teaming or security testing of AI systems
• Ability to work with frontier AI companies and government stakeholders

Responsibilities

• Lead a red team focused on stress-testing AI control measures
• Develop algorithms to find failures in control measures using adversarial ML
• Partner with frontier AI companies to assess and strengthen controls
• Bring understanding of control measures to government for policy decisions