Nuclear Safeguards Red Team Expert (Remote)
Listing checked September 15, 2026 · pay as published by Mercor
Overview
Neon seeks nuclear domain specialists to pressure-test frontier AI models on misuse judgment. You will craft single-turn prompts across three bands: benign, dual-use, and adversarial. Then you evaluate model replies against a policy standard and write the reference answer with your reasoning. The panel values fuel-cycle depth, safeguards, and nonproliferation work. The task-based remote engagement pays a one-time $75 fee.
What You'll Do5
- 1Write single-turn prompts in your nuclear specialty, tagging each as benign, dual-use, or adversarial.
- 2Test whether a frontier model answers a legitimate technical question in full while refusing a dangerous request.
- 3Score every model response against a defined policy standard and explain your fit judgment in writing.
- 4Draft the reference answer: what a correct reply includes and the technical basis behind it.
- 5Read and write about misuse scenarios in your field for long sessions. Take breaks whenever needed.
Requirements8
- 1Deep fuel-cycle knowledge plus exposure to how nuclear material is controlled, accounted for, or diverted.
- 2Safeguards or nonproliferation experience is the strongest signal.
- 3Red-teaming or adversarial prompt-writing background is a plus.
- 4Direct experience in one or more of these areas: IAEA or Euratom inspection, material accountancy, Additional Protocol, enrichment, reprocessing, criticality, MPC&A, insider threat, physical protection design, radioxenon monitoring, attribution, export control, or NSG trigger lists.
- 5Strong technical writing. Prior published research, expert witness work, or a writing sample you can share.
- 6Ability to explain a judgment to a non-specialist reader.
- 7No classified, export-controlled, NDA-covered, or prepublication-reviewed material may appear in your work. Tell us about any such obligations so we can scope the task.
- 8This role is writing-intensive.
Who Should Apply
Ideal candidates bring fuel-cycle depth and a safeguards or nonproliferation background. You can write a prompt that looks academic until an expert sees what it asks for, then explain in plain language why a model response crossed a line. A published paper, expert witness report, or technical writing sample gives your application a strong boost. The role is less suitable if you have broad nuclear policy knowledge but no fuel-cycle or material accountancy experience. The panel rejects applications that skip the writing sample or that base prompt examples on controlled or classified details.
Salary Insight
Pay is a one-time $75 fee. The amount is fixed for the task, not a per-hour or retainer rate, so clarify the expected number of prompts and review rounds before you accept. For expert red-team work with specialized nuclear knowledge, a one-time fee at this level often matches a short scoped contribution rather than a long project.
Location
Compensation
$75 fixed
Required Skills
Application Tip
Attach a technical writing sample, published paper, or expert witness report and connect it to one requirement from this post, such as safeguards, material accountancy, or fuel-cycle analysis. State any NDA, export-control, or prepublication obligations in your application so the panel can scope the work before you begin.
See NearSkill jobs more often in your search
Application & verification flow
1Instant rubric match
Your resume is scanned against this role’s requirements to check qualification fit.
2Screened before the employer sees it
Only profiles that clear screening are passed on.
3Outcome by email
We notify you at the address on your resume once the screening is reviewed.
Similar open positions
Explore active roles that match your skills and interests.
Mercor
VerifiedNuclear Security Engineer (Remote, Task-Based)
A panel of nuclear materials and safeguards specialists will red-team frontier AI models for Mercor, testing how well a model judges misuse potential. You write single-turn prompts at three levels: benign, dual-use, and adversarial, then judge model replies against a policy standard and draft the reference answer with technical reasoning. The work sits on the dual-use line where a material-balance calculation can close an inspector's account or hide a gap. Expect sustained reading and writing about misuse scenarios, with briefings and freedom to pause without penalty.
Mercor
VerifiedIAEA Safeguards Analyst (Remote, Task-Based)
The project brings together nuclear materials and safeguards experts to red-team frontier AI models. Your job is to write single-turn prompts that test whether a model can separate legitimate technical requests from dangerous ones. Prompts fall into three buckets: benign, dual-use, and adversarial. You then score model responses against a policy standard and draft the reference answer with the technical reasoning behind it. The work sits on the dual-use line that defines safeguards: a material-balance calculation can close an inspector's account or hide a diversion, and a fuel-cycle step can power a reactor or feed a proliferation pathway.
Mercor
VerifiedNuclear Fuel Cycle Engineer - Remote Task-Based
A task-based remote role puts you on a panel of nuclear fuel cycle and safeguards specialists who red-team frontier AI models. The core question is whether a model can tell a legitimate technical request from a misuse request, answering safe questions in full while refusing dangerous ones. You write single-turn prompts across benign, dual-use, and adversarial levels, then judge model replies against a set policy. Because fuel-cycle work sits on the dual-use line, the panel needs practitioners who know which parameters are routine engineering and which point toward proliferation. Your written reference answers show what a correct response should look like and why.
Mercor
VerifiedRadiological Safety AI Red Team Expert (Remote)
Neon seeks radiological safety practitioners to red-team frontier AI models on a task-based engagement. You will craft single-turn prompts that sit on the dual-use line, label each one as benign, dual-use, or adversarial, then judge how the model handles the request against a set policy. The work matters because a useful model must answer routine dose, shielding, or isotope questions while refusing requests that seek misuse details. Your domain knowledge decides where that boundary falls, and your writing makes the call clear to people outside radiological safety.
Mercor
VerifiedNonproliferation Analyst - Remote Contract
Neon is assembling a panel of nuclear materials and safeguards experts to red-team frontier AI models. The goal: see whether a model can separate legitimate technical questions from requests that enable misuse. You write single-turn prompts across benign, dual-use, and adversarial levels, then evaluate model answers against policy and draft reference responses with technical reasoning. This domain sits on the dual-use line, where an export control list can guide a legitimate shipment or reveal what a procurement agent should avoid declaring. The panel needs practitioners, not generalists, to draw that line with consistency.
Mercor
VerifiedForensic Scientist Nuclear Materials Remote
Neon is assembling a panel of nuclear forensics and safeguards practitioners to red-team frontier AI models. The job is to test whether a model can judge misuse potential: it should answer legitimate technical questions and refuse dangerous ones. You will write single-turn prompts across three labels, benign, dual-use, and adversarial, then assess model responses against a set policy standard. Each prompt also needs a reference answer that describes a sound reply and the technical reasoning behind it. Much of the challenge comes from the dual-use nature of this field, where the same signature that helps attribute a sample could also reveal what to hide.