Frontier AI Offensive Security Engineer

Aisi; Greater London; On-site; GBP 90,000 - 150,000; Full time

Job summary

AI Security Institute in London is seeking an experienced offensive security engineer to build and break across cloud, infrastructure and application layers. You will run engagements against evaluation sandboxes and the open-source Inspect framework, and develop agent-assisted red-teaming harnesses for continuous evaluation.

You will work with engineers and researchers on cyber ranges, publish findings, and present at talks. This role offers direct impact on frontier AI safety and governance.

Qualifications

• Substantial hands-on offensive security experience in red teaming, penetration testing or exploit development.
• Depth in at least two of: cloud/infrastructure (AWS preferred); container, VM or sandbox escape; application security; vulnerability research and exploit development.

Responsibilities

• Attack evaluation sandboxes and the environments that contain frontier models during evaluations, including container/VM escape and egress controls.
• Find vulnerabilities in Inspect and coordinate fixes and disclosure with maintainers and downstream users.
• Build and run agent-assisted red-teaming harnesses end to end, and test them against new sandbox/evaluation infrastructure before releases.
• Triages and proves findings, distinguishing real issues and validating exploits.
• Run agent and human red-teaming against targets, feeding insights back into tooling.
• Plan and run engagements across cloud, identity, model-access paths and research platform, including joint exercises with the detection team.
• Collaborate with researchers to build and harden cyber ranges, review evaluation tasks for realism, and provide human baselines.
• Oversee external testing scoping andReviews of penetration tests.
• Operate within governance and approval processes for model-assisted offensive work.
• Share work through publications, blogs and talks on sandbox security and AI-assisted offensive work.

Skills

Offensive security

Red teaming

Cloud and infrastructure security

Python/Go/Rust

Vulnerability research

Tools

Inspect framework

Sandbox/VM testing

Penetration testing tooling

Job description

AI Security Institute in London is seeking an experienced offensive security engineer to build and break across cloud, infrastructure and application layers. You will run engagements against evaluation sandboxes and the open-source Inspect framework, and develop agent-assisted red-teaming harnesses for continuous evaluation.

You will work with engineers and researchers on cyber ranges, publish findings, and present at talks. This role offers direct impact on frontier AI safety and governance.