Hiring nowAI application security and misuse defencePart of The agent workforce

Agent Security Engineer

Also called AI AppSec Engineer · Prompt Injection Defence Engineer

An Agent Security Engineer protects AI agents from prompt injection, data leaks and tool misuse. They let a company ship agents that act on real systems and customer data without handing attackers a new way in.

What they do

They map how an agent can be attacked, then close the gaps. That starts with threat modelling: what the agent can read, what it can call, and who can influence its input. They write and run attacks such as injected instructions in documents, web pages and tool output. They review tool permissions, limit what each agent can reach, and add checks before risky actions. They build monitoring to spot exfiltration and abuse. When something goes wrong, they lead the investigation and fix the cause, not just the symptom.

What a strong hire delivers

In the first months, a strong hire produces a clear threat model for the main agent features and a ranked list of real weaknesses. They tighten tool permissions and data access, and set up repeatable attack tests that run before each release. They put in place detection and an incident process the team has actually rehearsed. Engineers learn what safe agent design looks like.

How you can tell they are good

Ask them to walk through an attack they built or found against a model-connected system, and how they fixed it. Strong people explain why filtering prompts alone fails and talk about limiting capability, trust boundaries and data flow. Ask them to threat model a simple agent live. Be wary of answers that are only a list of jailbreak tricks.

Why it is hard to fill

Very few people have done this work for long, and the title varies: AI AppSec, prompt injection defence, agent security. Most candidates are strong in one half only. Classic AppSec people may not know how models behave. Researchers may lack production security habits. Every AI product company wants the same small group, so good people are rarely looking and often need a clear, serious mandate to move.

Where to find them

Most come from application security, threat detection or security research, then moved towards model-connected systems. Look at security engineers at AI product companies, red team members who have tested LLM features, and detection engineers who understand how agents call tools. Some come from vulnerability research or bug bounty work. Others come from platform security teams that reviewed early agent integrations.

Indicative pay

Not enough published evidence yet to give a fair range for this role, which is common for work this new. We benchmark it for you as part of a search.

Interview questions for an Agent Security Engineer

  1. Describe a prompt injection or tool abuse issue you found or fixed. What was the root cause?

    A strong answer is specific: the entry point, what the agent could reach, and the fix. Good candidates address capability and trust boundaries, not just wording. Worry if they only describe blocked phrases or cannot give a real example.

  2. How would you threat model an agent that reads email and can send messages and query a customer database?

    They should identify untrusted input, sensitive data and high-impact actions, then propose least privilege, confirmation steps and output controls. Look for attacker thinking and prioritisation. Worry if they list generic controls without tracing the data flow.

4 more questions, with what a strong answer covers, are in the Role Kit.

Job description you can post

We are building agents that act on behalf of our customers, and we need someone to make sure they cannot be turned against us or them. You will own how we find, test and fix security weaknesses in our model-connected products, working closely with engineers who build them.

What you will do

  • Threat model agent features, tools and data flows
  • Run red team exercises, including prompt injection and data exfiltration attempts
  • Review and restrict tool permissions and data access
  • Build automated attack tests that run before release
  • Design monitoring and detection for misuse and abuse
  • Lead investigation and fixes when incidents occur
  • Guide engineers on safe agent design

What you will bring

  • Demonstrated work securing applications, detection systems or research findings
  • Hands-on experience attacking or defending LLM-based systems
  • Strong understanding of how language models process instructions and context
  • Ability to read and write production code
  • Experience with tool and protocol integrations used by agents
  • Clear written communication of risk to non-specialists

Nice to have

  • Published research, advisories or bug bounty findings
  • Incident response experience in a production environment
  • Experience with agent frameworks

Free to use and adapt. If it saves you time, a link back to this page is appreciated.

Skills that matter

  • Security engineeringWorkingProtect systems, identities and sensitive assets
  • AI red teamingWorkingTest systems adversarially under authorisation
  • Tool and protocol integrationWorkingConnect models to external data and capabilities
  • Reliability and incident responseWorkingDetect failures, restore service and prevent recurrence

Other roles in aI application security and misuse defence

Hiring nowAI Security EngineerAn AI Security Engineer hardens the applications, interfaces and sensitive assets built around models. They find how attackers could misuse the…Hiring nowAI Red Team EngineerAn AI Red Team Engineer runs authorised attacks against model-connected systems to find weaknesses before real adversaries do. They give the company…Hiring nowAI Abuse Detection AnalystAn AI Abuse Detection Analyst spots unauthorised and harmful use of a company's models and products, then turns what they find into rules and cases…EmergingModel Supply Chain Security EngineerA Model Supply Chain Security Engineer protects model files, datasets and dependencies from tampering. They give the company confidence that what it…Hiring nowModel Weights Security EngineerA Model Weights Security Engineer protects trained model weights and the training infrastructure around them from theft, including by well-resourced…Hiring nowAI Threat Intelligence AnalystAn AI Threat Intelligence Analyst tracks how threat actors use AI to attack, deceive and abuse systems, then turns what they find into detections…EmergingAgent Identity and Permissions ArchitectThis person designs how AI agents prove who they are and what they may do. They keep access narrow as agent numbers grow, so one compromised agent…Hiring nowAI Trust and Safety SpecialistAn AI Trust and Safety Specialist writes and enforces the rules for how a model-connected product can be used, and investigates misuse. They keep…

Related roles in other families

Hiring nowCybersecurity ML ScientistCyber defence and security operations AIA Cybersecurity ML Scientist builds and tests models that spot malicious activity in authorised security data. They keep detection useful as…Hiring nowAI Detection EngineerCyber defence and security operations AIAn AI Detection Engineer builds the detections that turn AI-derived signals into alerts a security team can trust. They cut noise, catch real…EmergingAI Security Operations EngineerCyber defence and security operations AIAn AI Security Operations Engineer builds constrained AI support for security triage and response. It lets a security team move faster on alerts and…EmergingDefensive AI Agent EvaluatorCyber defence and security operations AIA Defensive AI Agent Evaluator tests how AI agents behave inside security detection and response work, and shows where they fail. It matters because…

How we wrote this

Role definitions are our own editorial work, checked against occupational classifications, standards, research and employer pages. Titles vary by company, and this is a description of the work, not a job advert.

← AI application security and misuse defenceAll new job families

Start a search

Hiring an Agent Security Engineer?

Tell us the work. We find the people already doing it, whatever their job title says, and you meet them already interviewed for your role.