Refractal is an AI security company building the infrastructure organisations need to defend against autonomous AI threats. 2026 has shown that generally capable intelligence is here. The challenge now is modernising security to stop AI-enabled threats, whether they come from internal agents going rogue or from attackers using AI to become more dangerous.
Refractal was founded on cybersecurity and AI expertise from MIT, Microsoft, NASA and the U.S. Navy. Our product lets organisations detect AI-enabled threats at low cost without sacrificing detection quality. We take a practical approach to AI risk, combining proven cybersecurity methods with frontier AI security research.
Today, we work with governments and organisations that can't afford to get security wrong. We're backed by leading deep-tech and cybersecurity investors and are growing the team to deliver on this mission.
The roleYou'll test how deployed AI systems and agents fail under adversarial pressure. You'll run authorised, controlled security assessments, show what each finding means in practice, and work with our engineering and research teams to turn findings into stronger defences and repeatable tests.
What you'll doDesign and execute assessments of AI applications and agents, grounded in realistic attacker goals and clearly defined test scope.
Test realistic threats such as indirect prompt injection, misuse of connected tools, sensitive data exposure and attempts to cross permission boundaries.
Build reproducible test cases and small testing tools that demonstrate security impact and distinguish reliable failures from one-off outputs.
Analyse how model behaviour, application design, identity and permissions combine to create vulnerabilities.
Document findings, recommend practical mitigations and retest fixes with the engineers responsible for the affected systems.
You've established a repeatable assessment approach that connects realistic adversaries, test scenarios and security outcomes.
You've produced actionable findings with reproducible evidence and helped validate the resulting fixes.
You've built a growing set of regression tests that captures recurring failure modes and checks whether fixes hold.
Hands-on experience in offensive security, application security testing or adversarial testing of AI systems.
A solid understanding of web applications, APIs, authentication, authorisation and common security failure modes.
An understanding of how AI applications use prompts, retrieved information, memory and connected tools, and where trust boundaries arise.
The ability to write code that automates testing, reproduces findings and inspects application behaviour.
Good judgement around test scope and sensitive data, and the ability to explain technical findings in terms of practical impact.
Experience assessing LLM applications, agent workflows or systems that retrieve external content.
Experience building security evaluation suites or testing proposed mitigations systematically.
Experience combining application testing with cloud, identity or infrastructure security assessment.
Radical transparency. We believe the best work happens when everyone feels able to give and receive honest, constructive feedback.
Public service. We see AI security as a form of public service. Getting the AI transition right means putting the right infrastructure in place to manage the threats that come with it, and our work helps protect the institutions and services society depends on.
Fail fast. We test ideas early, seek feedback and learn quickly. We experiment rapidly and change course when something isn't working, without compromising on security.
Pragmatism, not dogmatism. We choose approaches based on how well they solve the problem. That means combining established cybersecurity methods with frontier AI research, questioning our assumptions, and changing our minds when the evidence changes.
Compensation and benefitsSalary: £65,000–£125,000
Competitive equity to share in Refractal's growth
Pension contribution
Private medical and dental cover
Visa and relocation support
Skills Required
- Hands-on experience in offensive security, application security testing, or adversarial testing of AI systems
- Solid understanding of web applications, APIs, authentication, authorization, and common security failure modes
- Understanding of prompts, retrieved information, memory, connected tools, and trust boundaries in AI applications
- Ability to write code that automates testing, reproduces findings, and inspects application behavior
- Good judgment regarding test scope and sensitive data
- Ability to explain technical findings in terms of practical impact
- Experience assessing LLM applications, agent workflows, or systems retrieving external content
- Experience building security evaluation suites or systematically testing proposed mitigations
- Experience combining application testing with cloud, identity, or infrastructure security assessment
What We Do
Refractal builds security infrastructure for high-assurance AI, helping governments and enterprises test AI agents before deployment, protect them in production, and meet compliance requirements. Its software monitors agent activity and enforces policies at the point of action, evaluating tool calls and data access and recording tamper-evident audit trails. The platform integrates with existing telemetry and observability tools and helps detect AI-enabled threats across models, tools, and workflows.
.png){kind=logo}
