AI Security Researcher

CommonAI CIC; Cambridge; On-site; GBP 70,000 - 110,000; Full time

Benefits offered by this job

Competitive salary

Pension

Cambridge office near train station

Professional development

Networking opportunities

Job summary

CommonAI CIC is seeking an AI Security Researcher with an attacker mindset to help secure and audit AI systems used in regulated environments. You will track advances in model attacks, defenses, and security of serving pipelines while contributing to the High Assurance programme.

You will evaluate AI systems, test interpretability approaches, and share findings with Members. The role emphasizes threat modelling, red-teaming, and developing robust, auditable AI solutions within a collaborative

Qualifications

• Proven security research experience including offensive testing.
• Understanding of how modern AI systems train, deploy and serve.
• Familiar with AI attacks: jailbreaks, data poisoning, prompts.
• Strong Python and ability to build tools
• Excellent written communication for briefings
• Ability to challenge technical decisions constructively

Responsibilities

• Track key research and developments relevant to the High Assurance programme, including attacks on AI models, defences against them, and the security of the systems that serve them
• Report on what these developments mean for the programme and our Members, through written briefings and presentations
• Review the programme's technical decisions and directions from an attacker's point of view, and propose stronger alternatives
• Threat-model AI systems used in regulated settings, covering model weights, training data, inference serving, and the tools and agents built on them
• Red-team the systems we build, within an agreed scope, and report findings through a clear disclosure process
• Help develop interpretability methods for the simpler models produced by our frontier research, and test their limits, for example by trying to hide behaviour the methods should detect
• Build internal expertise by sharing what you learn with colleagues and helping answer technical questions from our Members

Skills

Security research

Offensive security

Adversarial ML

Python

Technical writing

LLM tooling

Attacker mindset

Education

PhD in security or ML

Master's in related field

Tools

Jupyter

PenTest tools

Git

Job description

About CommonAI CIC

CommonAI CIC is a non-profit membership organisation, founded on a belief in collaborative engineering for the safe and responsible development of foundational AI technologies. A place where AI startups, enterprises large and small, public sector bodies and academia can share resources and knowledge, to codevelop and grow businesses, fast.

About CommonAI CIC

CommonAI CIC is a non-profit membership organisation, founded on a belief in collaborative engineering for the safe and responsible development of foundational AI technologies. A place where AI startups, enterprises large and small, public sector bodies and academia can share resources and knowledge, to codevelop and grow businesses, fast.
We are led by experienced founders, investors and engineers who believe that collaborative engineering drives faster AI innovation and are backed by a mix of UK Government and private funding in order to design, build and deploy innovative AI systems.

The Opportunity

We are seeking an AI Security Researcher with the mindset of an attacker: someone who enjoys finding out how systems can be broken, misused or circumvented. Our research team works on reducing the complexity of large language models while retaining their performance.
You will support our High Assurance programme, which works on the accuracy, explainability and accountability of AI systems used in regulated environments such as financial services. You will track the research that matters to the programme, challenge its technical decisions, and help test whether the simpler, more interpretable models emerging from our frontier research are easier to audit in practice.

What You'll Do

• Track key research and developments relevant to the High Assurance programme, including attacks on AI models, defences against them, and the security of the systems that serve them
• Report on what these developments mean for the programme and our Members, through written briefings and presentations
• Review the programme's technical decisions and directions from an attacker's point of view, and propose stronger alternatives
• Threat-model AI systems used in regulated settings, covering model weights, training data, inference serving, and the tools and agents built on them
• Red-team the systems we build, within an agreed scope, and report findings through a clear disclosure process
• Help develop interpretability methods for the simpler models produced by our frontier research, and test their limits, for example by trying to hide behaviour the methods should detect
• Build internal expertise by sharing what you learn with colleagues and helping answer technical questions from our Members

Requirements

• Proven experience in security research, offensive security or adversarial machine learning (e.g. red-teaming, penetration testing, vulnerability research, or attacks on ML models)
• A strong understanding of how modern AI systems work, including how large language models are trained, deployed and served
• Familiarity with attacks on AI systems (e.g. jailbreaks, prompt injection, data poisoning, backdoors, model extraction)
• Strong Python, and the ability to build your own tools and experiments
• Excellent written communication: you can turn complex technical findings into clear briefings for technical and non-technical readers
• The judgement and confidence to challenge technical decisions constructively
• A strong enthusiasm for using LLMs, coding assistants, and agents as core tools in your own software development process

We also value:

• Experience with interpretability methods (e.g. probing, sparse autoencoders, circuit analysis), or a strong interest in learning them
• Experience working on software systems that operate in highly regulated or high-assurance environments (e.g. financial services)
• An understanding of the latest AI safety research and active involvement in that community
• Knowledge of hardware and side-channel security (e.g. timing attacks, risks in shared GPU and inference infrastructure)
• Familiarity with AI assurance standards and regulation (e.g. ISO/IEC 42001, NIST AI RMF, the EU AI Act)
• A public track record, such as published research, CVEs, conference talks, bug bounties or CTF results
• A PhD in security, machine learning or a related field, or equivalent research experience

Benefits

• A collaborative and supportive work environment
• The opportunity to have a high impact in a growing organisation
• Competitive salary package and pension
• Professional development opportunities
• Networking opportunities with influential people from across the tech sector, financial services, and academia
• A vibrant office environment located a few minutes' walk away from Cambridge train station

CommonAI CIC is an equal opportunity employer and is committed to creating an inclusive and diverse workplace.