Radiologicals Safety for Redteaming

The rapid evolution of artificial intelligence has propelled frontier models into domains previously reserved for highly specialized human experts. As AI systems become more capable across technical, medical, and industrial sectors, ensuring their safety and reliability is paramount. Tech labs and research groups are no longer relying on generic content moderation; they need world-class professionals who understand the granular realities of high-hazard disciplines.

Mercor is currently spearheading an elite recruitment initiative to assemble a specialized panel of radiological safety experts for the Radiologicals Safety for Redteaming project. Offering lucrative compensation ranging from $65 to $75 per task, this fully remote opportunity allows certified professionals—such as Radiation Safety Officers, Health Physicists, and Source Security Specialists—to directly shape how advanced language models handle sensitive, dual-use information.

Whether you are looking to leverage your extensive background in nuclear regulations, radiation protection, or emergency response for high-impact advisory work, this comprehensive guide covers everything you need to know about the role, the responsibilities, and how to successfully complete your application.

The Core Challenge: Navigating the Dual-Use Line in AI

In the realm of radiological science, the line between beneficial application and dangerous misuse is often razor-thin. Consider radioactive isotopes: the exact same isotope underpins life-saving cancer therapy and vital industrial radiography. Similarly, standard shielding calculations protect workers in a clinical environment while also governing the physical security required to conceal a source.

For artificial intelligence models, recognizing this delicate distinction is exceptionally difficult. A model that reflexively refuses a routine, professional dosing question is failing just as severely as one that answers a hazardous prompt it has no business addressing. Drawing that boundary reliably requires deep, hands-on practitioner experience rather than surface-level generalist knowledge.

As an expert red-teamer on this project, your primary objective is to test, stress-test, and refine how frontier models navigate these exact gray areas.

What You Will Do on the Project:

• Drafting Challenging Prompts: Write intricate, single-turn prompts rooted in your specific domain expertise, categorized across three distinct levels: benign, dual-use, and adversarial.
• Rigorous Model Evaluation: Critically evaluate the responses generated by AI models against established safety policy standards, determining whether each output was handled correctly.
• Developing Reference Answers: Author comprehensive reference answers outlining what a correct, safe response looks like, accompanied by detailed technical reasoning for why the model succeeded or failed.

Target Backgrounds and Qualifications

Mercor is not looking for casual writers or generic tech workers for this panel. They are seeking proven practitioners who have held legal, administrative, or operational responsibility for radioactive material and high-hazard environments. Ideal candidates typically possess experience in one or more of the following specialized tracks:

• Radiation Safety Officer (RSO): Broad-scope licensees who have held the official license of record, managing material acquisition, inventory tracking, secure transfers, and final disposal.
• Health Physicists: Professionals from Department of Energy (DOE) sites or national laboratories with direct experience in radiological surveys, assaying, contamination control, and dose planning in high-hazard settings.
• Source Security & Vulnerability Assessment Specialists: Individuals who have professionally evaluated how Category 1 and 2 sources are secured, transported, or shielded against security vulnerabilities.
• Radiological Emergency Response Specialists: Experts versed in Federal Radiological Monitoring and Assessment Center (FRMAC) protocols, CBRN response, and contamination dispersal modeling.
• Nuclear Medicine Physicists & Cyclotron RSOs: Specialists focused on medical isotope production, handling, and the complex clinical mechanics of the dual-use spectrum.

Beyond Technical Depth: The Writing Demand

Because this project involves evaluating complex AI logic, strong writing skills are essential. Every evaluation and judgment you log must include a clear, cohesive written rationale that a non-specialist program manager can easily follow. Candidates with a background in technical writing, published academic research, or expert witness testimony will find themselves exceptionally well-suited for this engagement.

Compensation Structure and Professional Perks

Mercor’s contract model is designed for high-level experts who value autonomy, efficiency, and competitive remuneration:

• Task-Based Earnings: Earn between $65 and $75 per completed task, allowing you to scale your income based on your output and availability.
• Absolute Flexibility: Complete tasks entirely on your own schedule from any remote location. You can pause or step away from sensitive subject matter reading at any point without penalty.
• Lucrative Referral Opportunities: Mercor offers expansive peer referral bonuses—allowing you to earn up to $300 for each successful specialist referral who joins the platform.

Step-by-Step Guide to Acing Your Application

Because positions requiring high-clearance domain expertise are highly competitive, presenting a clean, focused application is essential:

1. Highlight Your Hands-On Credentials: Explicitly detail your certifications, license holdings (such as broad-scope RSO credentials), or institutional affiliations on your profile.
2. Provide Writing Samples: If you have published papers, technical safety documentation, or public safety reports, link them or attach samples to showcase your written clarity.
3. Commit to the Screening Process: Complete Mercor’s streamlined interview and assessment steps promptly to secure early placement on the red-teaming panel.

AI safety labs are actively expanding their red-teaming rosters. Secure your position today and help define the safety boundaries of next-generation intelligence.