EmergingAI control, monitoring and containmentPart of The agent workforce

AI Control Engineer

Also called Control Protocols Engineer · Monitoring Engineer

An AI Control Engineer builds the monitoring, sandboxing and containment that let a company use powerful models safely, even if a model is misaligned or trying to get around the rules. It limits harm when trust cannot be assumed.

What they do

The work starts from an awkward assumption: the model might be working against you. The engineer designs protocols that keep it useful anyway. That means building monitors that review model actions, setting up sandboxes with tight permissions, and deciding when an output needs a second check or a human. They red-team their own setups, playing the part of a model that tries to slip bad actions past oversight. They measure how often controls fail, tune thresholds, and write up the weak points. Much of the day is systems and security engineering, with evaluation work alongside it.

What a strong hire delivers

In the first months, a strong hire maps where model actions can cause harm and puts working controls around the riskiest ones. They stand up a sandbox and a monitoring pipeline that the team actually uses. They run an adversarial test of those controls and report honestly on what got through. They leave behind clear threat models, measurable results and a short list of next fixes.

How you can tell they are good

Ask them to walk through a control setup they built and then attack it. A strong candidate names the threat model first, explains what the monitor cannot catch, and shows how they measured failure. Ask how they would handle a model that behaves well in testing. Worry if they treat alignment as a substitute for containment, or only talk in theory.

Why it is hard to fill

Very few people have done this work under this title. It sits inside security, evaluation and safety teams, so candidates rarely look for it by name. The mix is unusual: security instincts, ML understanding and systems skill in one person. Labs and government safety institutes draw on the same small pool, so searches take careful, direct outreach and a clear explanation of the problem.

Where to find them

Look first at security engineers who have moved toward ML systems, and at AI safety evaluation staff who build as well as test. Formal methods people are another source, especially those who have worked on verification of real systems. Many sit in frontier labs, AI safety institutes and small safety research groups, often under titles like monitoring engineer, security engineer or research engineer.

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 AI Control Engineer

  1. Describe a monitoring or containment system you built. What threat did it assume, and where did it fail?

    A strong answer states a clear threat model, names real failure modes and shows how they were found. Look for honest limits and iteration. Be wary of claims that the system was complete or never tested against an attacker.

  2. How would you test whether a monitor can be fooled by a model trying to hide bad actions?

    Good answers describe red-teaming with an adversarial model or human, held-out attack strategies and measured miss rates. They consider the monitor being weaker than the model. Worry if testing only uses benign or obvious examples.

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

Job description you can post

We are hiring an AI Control Engineer to build the monitoring, sandboxing and containment that let us use powerful models safely, even if a model behaves in ways we do not trust. You will work across security, evaluation and systems, and your work will shape what we allow our models to do.

What you will do

  • Design and build monitoring for model actions and outputs
  • Build and maintain sandboxes and permission limits for model agents
  • Define threat models for untrusted or adversarial model behaviour
  • Red-team our own controls and report what gets through
  • Measure control failure rates and tune escalation thresholds
  • Work with research and product teams to set safe deployment conditions

What you will bring

  • Built security-sensitive systems such as sandboxes, access controls or audit pipelines
  • Practical experience evaluating model behaviour
  • Able to write a clear threat model and test against it
  • Strong engineering skills in production systems
  • Comfortable reasoning about adversaries who adapt
  • Clear written communication of risk and uncertainty

Nice to have

  • Experience with formal verification
  • Background in AI safety evaluation or red-teaming
  • 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

  • AI controlWorkingLimit harm from potentially untrusted AI behaviour
  • Security engineeringWorkingProtect systems, identities and sensitive assets
  • Model evaluationWorkingMeasure capability and failure modes credibly
  • Formal verificationWorkingProve stated properties under explicit assumptions

Other roles in aI control, monitoring and containment

EmergingAI Control ResearcherAn AI Control Researcher designs the safeguards that limit what a model can do if it turns out to be unreliable or hostile. They keep risk bounded…EmergingAgent Containment EngineerAn Agent Containment Engineer builds the boundaries that keep AI agents from doing harm, even if the model misbehaves or is attacked. They matter…EmergingAI Monitor Evaluation EngineerAn AI Monitor Evaluation Engineer tests whether the systems that watch model behaviour actually catch dangerous or unauthorised actions. It matters…EmergingAI Shutdown and Recovery EngineerAn AI Shutdown and Recovery Engineer builds and tests the ways to stop an AI system mid-action and bring it back to a known good state. It matters…

Related roles in other families

EmergingAI Formal Verification EngineerFormal methods and reliable reasoningAn AI Formal Verification Engineer writes precise specifications for AI systems and their outputs, then proves they hold or finds counterexamples…Hiring nowAutomated Theorem Proving ResearcherFormal methods and reliable reasoningAn Automated Theorem Proving Researcher builds and improves methods that find or check mathematical and logical proofs by machine. The company gets…Hiring nowProgram Synthesis ResearcherFormal methods and reliable reasoningA Program Synthesis Researcher builds systems that generate programmes which meet a testable or formal specification. They give the company code…EmergingAI-Assisted Verification SpecialistFormal methods and reliable reasoningAn AI-Assisted Verification Specialist uses AI tools to speed up proof and specification work, then checks every result independently. They help a…

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 control, monitoring and containmentAll new job families

Start a search

Hiring an AI Control 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.