EmergingAI biosecurity and scientific misuse assessment

Scientific Misuse Safeguards Researcher

Also called Science AI Safeguards Researcher

This person tests whether an AI system's safeguards hold when it is asked for high-consequence scientific help. They give a company evidence that its controls reduce misuse risk without blocking legitimate scientific work.

What they do

They design and run authorised adversarial evaluations against a model's safeguards in the sciences, working in secured settings with agreed rules. They map plausible misuse routes, build test sets, and probe classifiers, refusals and access controls. They read results with care, separate real weaknesses from noise, and write up what failed and why. They work with safety, policy and engineering teams to adjust the safeguards, then retest. They also measure how often a safeguard blocks genuine researchers, and report that cost alongside the risk reduction.

What a strong hire delivers

In the first months, a strong hire delivers a documented safeguard evaluation method the team can repeat. They run an authorised adversarial evaluation and report the findings clearly to decision makers. They agree how risk reduction and false positives will be measured. They leave behind a prioritised list of gaps, fixes that have been retested, and a review process that engineers and risk owners will actually follow.

How you can tell they are good

Ask them to walk through an evaluation they designed, including its limits. Good candidates explain how they chose threat scenarios, how they handled sensitive material, and how they judged whether a safeguard worked. They talk about false positives without prompting. They can explain science to engineers and evaluation results to scientists. Be wary of anyone who treats refusal rates as the only measure.

Why it is hard to fill

The work sits between biosecurity, life sciences and AI evaluation, and few people have all three. Many are doing it now under titles such as red teamer, safety researcher or biosecurity analyst. Trust matters, as candidates handle sensitive material and need to be comfortable with controlled access and vetting. Strong scientists often lack evaluation experience. Strong evaluators often lack domain depth.

Where to find them

Look in biosecurity and biosafety teams, public health and life sciences research, and groups that assess dual-use risk. Also look at model evaluation, red teaming and trust and safety teams at frontier labs, model providers and safety-focused research organisations. Some come from government or academic risk assessment. The best fits often have one deep field and real hands-on evaluation work in another.

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 a Scientific Misuse Safeguards Researcher

  1. Describe an adversarial evaluation you designed or ran. How did you decide what to test and when to stop?

    Covers threat scenarios, scope, authorisation, coverage and diminishing returns. Explains what was found and what changed. Worry if they describe only ad hoc probing, cannot say what was left untested, or ignore how findings were acted on.

  2. How do you measure whether a safeguard reduces risk without harming legitimate scientific users?

    Names both sides: risk reduction and false-positive effects. Mentions representative benign queries, expert review and trade-offs. Worry if they rely on one number, or treat blocking everything as a safe default.

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

Job description you can post

We are hiring a Scientific Misuse Safeguards Researcher to test whether our AI systems give safe answers on high-consequence scientific topics. You will design authorised adversarial evaluations, report what you find, and help us improve safeguards while keeping the system useful to legitimate researchers.

What you will do

  • Design and run authorised adversarial evaluations of safeguards in high-consequence scientific areas
  • Develop and document a repeatable safeguard evaluation method
  • Analyse misuse threats and turn them into test scenarios
  • Measure risk reduction and the false-positive effects on legitimate users
  • Report findings clearly to engineering, policy and risk teams
  • Retest fixes and maintain risk-review processes
  • Handle sensitive material under agreed controls

What you will bring

  • Demonstrated work in biosecurity, biosafety, life sciences or authorised model evaluation
  • Experience designing or running red teaming or evaluation of AI systems
  • Ability to analyse misuse threats with sound judgement
  • Clear written reporting that non-specialists can act on
  • Careful handling of sensitive information
  • Comfort working with engineers and scientists alike

Nice to have

  • Experience in risk management or dual-use review
  • Experience with secured evaluation environments
  • Background working with classifiers or access controls

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

Skills that matter

  • Scientific domain expertiseAdvancedApply disciplinary theory and experimental standards
  • AI red teamingAdvancedTest systems adversarially under authorisation
  • AI risk managementWorkingIdentify, assess, treat and monitor AI risks
  • Model evaluationWorkingMeasure capability and failure modes credibly

Other roles in aI biosecurity and scientific misuse assessment

EmergingAI Biosecurity EvaluatorAn AI Biosecurity Evaluator tests whether a model gives meaningful uplift towards biological harm, under secured conditions. They give the company…EmergingAI Biosafety SpecialistAn AI Biosafety Specialist reviews how AI workflows touch laboratory and biological work, and puts protective procedures in place. They help a…EmergingAI Biosecurity Policy AnalystAn AI Biosecurity Policy Analyst weighs the options for managing biological risk from AI systems and sets out what is feasible. They help a company…

Related roles in other families

Hiring nowAI Robustness EngineerAI robustness, safety engineering and incidentsAn AI Robustness Engineer tests how a model behaves when inputs change or the world shifts, then builds and checks fixes. It matters because…Hiring nowAI Safety EngineerAI robustness, safety engineering and incidentsAn AI Safety Engineer turns the ways an AI system could hurt users or the business into specific tests and controls. That gives the company evidence…EmergingAI Incident Response LeadAI robustness, safety engineering and incidentsAn AI Incident Response Lead coordinates the response when a deployed AI system fails or causes harm. They contain the damage, find the root cause…EmergingAI Reliability Evaluation ScientistAI robustness, safety engineering and incidentsAn AI Reliability Evaluation Scientist measures how consistently an AI system works and how likely it is to fail on realistic workloads. It gives…

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 biosecurity and scientific misuse assessmentAll new job families

Start a search

Hiring a Scientific Misuse Safeguards Researcher?

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.