You and the process

With a person
Anna, in-house recruiter
9 years hiring engineers
30 minutes with a real recruiter
They read your CV with you, on a call, and say where the offers are being lost.
Didn’t find what you were looking for? Tell us what to build
Early Window: Closes in
Company hidden

Research Residency

  • Hybrid
  • 0-3 years

Salary

£65,000/ year

AI summary

For members

The whole posting in a few lines. Sign up to read it here and on every role you open.

Sign Up to Read

Description

It sits within the UK government, with direct access to the Prime Minister's office, and works with frontier AI developers and governments around the world.

The Red Team tests the safety measures of frontier AI systems and shares its findings with leading AI companies, UK officials and allied governments. Demand for this work is outpacing current capacity, so the team is growing.

The team has three sub-teams. Alignment builds methods to check whether models reliably act as intended. Control studies how to stress-test measures that stop AI agents from taking harmful actions. Misuse tests safeguards against misuse of dangerous capabilities. Each sub-team has its own focus, but methods, infrastructure and tooling are shared across the whole team.

This is a 6-month residency by default, though the length can be shortened or extended to suit the candidate. Promising residents may be offered a full-time role in their preferred sub-team after an end-of-residency review.

Applicants should be PhD students, or postdoctoral researchers whose PhD was submitted within the past 2 years, in machine learning, computer science, computer security or a related field.

The selection process usually includes an initial assessment, a screening call, a technical assessment, a research interview and a final interview with senior leadership. Some or all of these stages may apply, and the process can differ by candidate.

The deadline for applications is end of day Tuesday 8th December, anywhere on Earth.

Responsibilities

  • Work with research scientists to decide which questions matter most and how to test them.
  • Take on a defined project with support from an experienced research engineer or scientist. Projects may cover new attack methods, automated red teaming, alignment evaluations, or attacks on AI monitors and agents.
  • Help design and run attacks and evaluations on frontier AI systems to find weaknesses in their safety measures, such as jailbreaking safeguards, uncovering misaligned behaviour or evading control monitors.
  • Write code and build tools to run experiments at scale, for example testing attack techniques across several models, with a focus on reliability, speed and reproducibility.
  • Analyse experiment results carefully and turn them into clear findings for both technical and policy audiences.
  • Keep up with new attack techniques and research, and apply what is relevant to the team's work.
  • Learn how frontier AI models are evaluated and red teamed in practice.
  • Possible projects include building automated research methods for better red-teaming algorithms, such as evolutionary and multi-agent approaches. Misuse work may cover jailbreaking attack pipelines, backdoors and data poisoning, and prompt injection and agent security. Alignment work may cover automated detection of harmful or unsanctioned behaviour such as research sabotage or reward-seeking, more realistic simulated evaluations, and after-the-fact investigation of misaligned behaviour seen in the wild. Control work may cover better ways to evaluate AI monitors and the limits of control evaluations, both practical and fundamental. Your actual project may differ, as priorities change quickly and the team will try to match work to your interests and skills.

Requirements

  • Research experience in adversarial machine learning, AI security, alignment, AI control or evaluations, or familiarity with this research.
  • Experience with optimisation methods such as reinforcement learning, fine-tuning or evolutionary methods.
  • Familiarity with loss-of-control threat models, such as deceptive alignment, research sabotage or reward-seeking.
  • Hands-on experience red teaming or jailbreaking AI systems, or taking part in capture-the-flag, bug bounty or other security challenges.
  • Experience building something with LLMs, such as an evaluation, an agent or a fine-tuned model.
  • Experience using ML frameworks such as PyTorch or Hugging Face.
  • Public sharing of research or technical work, such as a paper, preprint, open-source tool or blog post.
  • Experience running experiments or analysing data, for example in coursework or projects.
  • Good research habits, such as keeping work reproducible, recording what you tried and being honest about uncertainty.
  • Curiosity about how systems fail, and willingness to try unusual approaches.
  • Clear written and spoken communication, including explaining technical work to non-specialists.
  • Openness to feedback, and willingness to ask for help when stuck.
  • Motivation to reduce risks from advanced AI and support government work on AI security.
  • Highly Desirable: published or co-authored research on adversarial ML, jailbreaks, data poisoning, AI safeguards, alignment, control or evaluations.
  • Highly Desirable: earlier research experience in ML or AI safety, such as a research internship, research assistant role or substantial thesis project.
  • Highly Desirable: a security research background, such as disclosing real vulnerabilities or placing well in CTF competitions.
  • Highly Desirable: experience fine-tuning or modifying open-weight models, or working with model internals.
  • Highly Desirable: familiarity with the alignment or AI control literature, or with current methods for post-training and aligning LLMs.
  • Highly Desirable: experience with RL, supervised fine-tuning or other ML optimisation, including careful choice of what to measure.
  • Highly Desirable: security analysis of agent scaffolds, sandboxes or ML infrastructure.
  • Highly Desirable: security testing of AI agents, such as prompt injection or tool misuse.
  • You do not need to meet every point. Apply even if you match only a few.
  • This role may not suit you if you want to set research strategy from day one, since residents work with experienced research leaders who set the agenda. It may also not suit you if you need well-defined requirements and long planning cycles, or if you prefer long stretches of solo work to collaboration.

Conditions

Salary is set by role scope and relevant experience. Most offers fall between £65,000 and £145,000, paid as a base salary plus a technical allowance, where take-home pay equals base plus technical allowance.

This role sits outside the DDaT pay framework because it needs deep technical expertise in frontier AI safety, robustness and advanced AI architectures.

Level 3: £65,000 to £75,000 (base £40,850 plus technical allowance of £24,150 to £34,150).

Level 4: £85,000 to £95,000 (base £48,540 plus technical allowance of £36,460 to £46,460).

Level 5: £105,000 to £115,000 (base £63,165 plus technical allowance of £41,835 to £51,835).

Level 6: £125,000 to £135,000 (base £76,475 plus technical allowance of £48,525 to £58,525).

Level 7: £145,000 (base £76,475 plus technical allowance of £68,525).

Successful candidates must pass a criminal record check and obtain baseline personnel security standard (BPSS) clearance before appointment. Eligibility for counter-terrorist check (CTC) clearance is strongly preferred. Some roles may need higher clearance, which will be stated in the advert where it applies.

Applicants from any nationality or background may be considered, so you can apply even if you do not meet the standard nationality requirements.

AI tools may help with your application, but all examples and statements must be truthful, accurate and drawn from your own experience. Presenting others' ideas or AI-generated text as your own may lead to withdrawal of the application or disciplinary action for internal candidates.

Civil servants dismissed for internal fraud, or who would have been dismissed had they resigned, are banned from civil service employment for 5 years and recorded on an internal fraud register checked during pre-employment screening.

The company supports diversity and equal opportunities, and runs a Disability Confident Scheme and a Redeployment Interview Scheme.

Benefits

  • Work alongside highly capable, mission-driven and supportive colleagues.
  • Influence how frontier AI is governed and deployed around the world.
  • Work with the Prime Minister's AI Advisor and leading AI companies.
  • Help shape what is described as the first and best-resourced public-interest research team focused on AI security.
  • Pre-release access to several frontier models and substantial compute.
  • Operational support that lets you focus on research and move quickly.
  • Work with experts in national security, policy, AI research and adjacent sciences.
  • Own important problems early if you are talented and driven.
  • Freedom to pursue research ideas without product pressure.
  • Opportunities to publish and collaborate externally.
  • Work from a modern central London office, or from government offices in Birmingham, Cardiff, Darlington, Edinburgh, Salford or Bristol where applicable. Hybrid working is available, with occasional remote work abroad and a stipend for home-office equipment.
  • At least 25 days of annual leave, 8 public holidays, extra team-wide breaks and 3 days off for volunteering.
  • Generous parental leave: 36 weeks of UK statutory leave shared between parents, 3 extra paid weeks, and an option for additional unpaid time.
  • Pension choice between Partnership, with no employee contributions and employer contributions of up to 14.75%, or Alpha, a defined benefit scheme giving guaranteed, inflation-linked retirement income for life.
  • Discounts and benefits for cycling to work, donations, and retail and gym memberships.
  • These benefits apply to direct employees. Secondees and others on different arrangements may receive different terms.

Where you’d work

Part of the week in the office

You can work from

  • United Kingdom

About the company

Company hidden

Office in London, United Kingdom

Your chances

Still hiring, not crowded yet, and you'd be among the first.

  • 17 checks run
  • 5 good signs
  • 1 red flag

Still hiring?

11 checks

Actively hiring

In its favour3

  • Still on the company's own careers site, checked 1 h agoModerate evidence
  • Found in the last 48 hours, before the big job boardsModerate evidence
  • States its salary: £65KSlight evidence

How crowded?

6 checks

Low

In its favour2

  • In its Early Window: not on the big job boards yetStrong evidence
  • Hybrid in London: only people nearby can take itSlight evidence

Against it1

  • Entry level: the most applied-to level there isModerate evidence

Fits Me

How well does this role fit you?

Answer a few questions or drop your CV, and every role gets a fit score with the reasons, this one first.

  • Your field
  • Level
  • Stack
  • Work model
  • Salary floor
  • Must-haves
Details10 facts · Role, Location, Compensation
Tech stack
  • PyTorch
  • LLMs
Specialty
Security
Region
United Kingdom
Pay period
Annual
Show 6 more factsShow less

Role

Category
DevOps & Infrastructure
Specialty
Security
Experience
2+ years
Tech stack
  • PyTorch
  • LLMs

Location

Work model
Hybrid
Region
United Kingdom
Office
  • London, United Kingdom
Remote from
  • United Kingdom

Compensation

Salary
£65,000 / year
Pay period
Annual

Something wrong with this vacancy?

Similar vacancies

Share this vacancy

What's wrong with it?

The employer never sees who reported.

Reason