Offensive Security Engineer
UK government research organisation that evaluates advanced AI risks and develops and tests mitigations to inform governments.
Funding history
Investors
About The AI Security Institute (AISI)
The AI Security Institute is a research organisation within the UK Department for Science, Innovation and Technology. It tests leading AI systems, conducts technical research on advanced-AI risks and solutions, develops evaluation infrastructure, and informs UK and allied-government policy.
Skills
Candidate Availability
Required and preferred rules are kept separate and reflect the wording in the original posting.
About the Role
You will conduct offensive security testing across evaluation sandboxes, cloud infrastructure, identity systems, applications, and research platforms. You will build and operate agent-assisted red-teaming harnesses, validate findings with working exploits, coordinate remediation and disclosure, support cyber-range development, manage external penetration tests, and share appropriate research publicly.
Requirements
- Substantial hands-on offensive security experience in red teaming, penetration testing, vulnerability research, or exploit development
- Depth in at least two areas: cloud and infrastructure, container/VM/sandbox escape, application security, or vulnerability research and exploit development
- Strong Python or another language such as Go, Rust, or C
- Experience writing and maintaining tools used by others
- Experience testing live systems safely
- Clear written and verbal communication and ability to drive remediation across teams
- Serious interest in frontier AI safety
- Successful completion of UK Government SC clearance
- Normally resident in the UK for the past 5 years
- Successful criminal record check and BPSS clearance
Responsibilities
- Attack evaluation sandboxes, including container and VM escape, egress controls, and evaluation hosts
- Find vulnerabilities in Inspect and its supporting infrastructure, and coordinate fixes and disclosure
- Build, operate, and maintain agent-assisted red-teaming harnesses
- Run automated and manual testing before sandbox and evaluation-infrastructure releases
- Triage agent findings and confirm genuine issues with working exploits
- Compare agent and human red-teaming results and improve tooling
- Plan and run scoped engagements against cloud, identity, model-access, and research-platform systems
- Help build and harden cyber ranges and review evaluation tasks
- Scope and review external penetration tests
- Follow internal safety governance and approval processes for model-assisted offensive work
- Publish research, write blog posts, and give talks where appropriate
Benefits
- Hybrid working
- Flexibility for occasional remote work abroad
- Work-from-home equipment stipend
- At least 25 days of annual leave
- 8 public holidays
- Extra team-wide breaks
- 3 volunteering days
- Paid parental leave
- Employer pension contribution of 28.97% of base salary
- Cycling, donation, retail, and gym discounts
- Learning and development stipends
- Conference and external collaboration funding
Hiring Process
Initial assessment → initial screening call → technical assessment → research interview → behavioural interview → final interview with members of the senior leadership team.