AI Security Institute (UK) Careers
The UK AISI is the largest single employer of AI safety and AI security talent inside any government. As of mid-2026 the institute had roughly 100 to 150 technical staff and was continuing to grow at a planned 30 to 50 new technical hires per year. It sits inside the Department for Science, Innovation and Technology (DSIT) but operates with substantial autonomy on hiring and research direction. The institute publishes its evaluations, partners directly with the frontier labs (Anthropic, OpenAI, Google DeepMind), and has become a key node in the global AI safety ecosystem.
What roles the UK AISI hires for
Research engineer. The largest single hiring category. Builds and runs evaluations of frontier AI models for dangerous capabilities, alignment failures, and security vulnerabilities. Strong applied ML engineering background expected. The Inspect framework (AISI's open-source evaluation library) is the technical centerpiece. 2026 compensation: £100,000 to £165,000 total ($125,000 to $205,000).
Senior research engineer. Leads evaluation programs, mentors junior staff, and contributes to AISI's research agenda. Typically requires 5+ years of strong ML engineering or equivalent experience. 2026 comp: £140,000 to £200,000 total ($175,000 to $250,000).
Research scientist. The research-track equivalent. PhD typically expected (in ML, statistics, or a strongly related field) with publications in alignment, evaluation, or AI security. 2026 comp: £145,000 to £215,000 total ($180,000 to $270,000).
AI evaluations specialist. A newer category created in 2024 specifically for people who design evaluation methodology. Sits between research engineer and research scientist in scope. Often filled by people with a mix of applied research and product engineering background.
Policy adviser. Translates AISI's technical findings into policy guidance for DSIT, UK government, and international AI governance bodies. Background varies (law, public policy, technical) but technical literacy is required to contribute meaningfully.
Operations and platform. Engineering management, platform engineering, security engineering, and operations roles that keep the institute running. Includes infrastructure security, which is one of the entry points for AI security engineers who do not have a research background.
UK AISI compensation in detail
UK AISI pay uses civil service banding with technical roles eligible for "specialist allowance" on top of the base civil service salary. The bands are wider than industry would suggest because the institute can pay near the top of each band for in-demand specialists. Bonuses are smaller than industry (typical 5% to 15% annual) and there is no equity. The pension is the UK Civil Service Pension Scheme, which is genuinely valuable (alpha scheme accrual) and worth modeling into total comp if you compare against industry offers.
The total-compensation gap to a frontier lab in London is significant. An Anthropic London research engineer at the same level earns roughly £230,000 to £350,000 total, against the AISI band of £100,000 to £200,000. People take the AISI role for mission, public interest, the chance to evaluate every major frontier model, and the unusually low political overhead inside a young institute. The pay cut is real and worth being honest about.
How to apply to the UK AISI
Two routes. Roles are posted on the AISI's own careers page (aisi.gov.uk/careers as of 2026) with application forms, and most senior roles also appear on Civil Service Jobs (civilservicejobs.service.gov.uk) where you apply through the standard civil service form. Both routes feed the same hiring loop.
The application form is more structured than industry applications. You will be asked for written responses against specific competencies (technical depth, communication, collaboration). These responses are scored against published civil service success-profile criteria, so write them concretely with specific examples. Vague answers fail at the sift stage even for strong candidates.
UK AISI hiring process and timeline
Typical timeline from application to offer is 6 to 12 weeks, with security clearance adding 2 to 4 months before a confirmed start date. The interview loop has three stages: an initial technical or written assessment (varies by role), a panel interview with 2 to 3 AISI staff (technical depth plus competency-based behavioral), and a final discussion that includes leadership. The bar is high and pass rates at each stage are low. Most successful candidates report 3 to 5 hours of interview time across the loop, with the technical assessment running another 2 to 4 hours.
Security clearance is required for most technical roles. The level varies but Counter-Terrorism Check (CTC) is the minimum and Security Check (SC) is common. Developed Vetting (DV) is rare but used for senior roles with the deepest access. Clearance is sponsored by AISI after a conditional offer. You do not need clearance to apply.
UK AISI vs US AISI vs frontier labs
The UK AISI typically pays more than the US AISI for the same role (the UK has structured the specialist allowance more flexibly than the US civil service has), but less than industry. The US AISI has different access constraints (US citizenship for most roles) but pays in dollars. The frontier labs pay multiples more but with industry trade-offs (less stability, more product pressure, equity risk). For senior research engineers specifically, the gap to a frontier lab is the steepest and the gap to a generic enterprise AI engineering role is the smallest. See AI safety jobs for the broader comparison and best AI security companies in 2026 for the industry alternatives.
Who actually gets hired at UK AISI
Three profiles dominate 2026 hires. Researchers from academic ML groups (Oxford, Cambridge, Imperial, UCL, plus international universities) who want to do safety-relevant work at scale. Engineers from industry (Google, Meta, DeepMind, Anthropic) who want public-interest impact. AI security specialists with offensive security backgrounds (penetration testers, red teamers, security researchers) who have started doing AI-specific work. The third profile is the fastest-growing because the institute has been expanding its security and evaluations capacity since 2024. If you are an AI security engineer with red-team experience, the UK AISI is a credible destination.
Get the AISec Brief
Weekly career intelligence for AI Security Engineers. Salary trends, who's hiring, threat landscape shifts, and certification updates. Free.