Nuclear Safeguards and AI Safety Evaluator

ixolabs.ai; France; Hybride; EUR 84 000 - 96 000; Plein temps

Résumé du poste

ixolabs.ai is seeking contractors to design and evaluate adversarial prompts within the approved nuclear-safety context. You will craft scenarios in safeguards and nonproliferation, then assess model reasoning for accuracy and compliance with criteria.

Applicants should have nuclear-sector experience, exposure to AI evaluation, and the ability to document explicit judgments and assumptions with reproducible checks. Remote assignments are possible with flexible scheduling.

Qualifications

• Professional nuclear-sector depth with safeguards or nonproliferation work.
• Experience in AI evaluation or red-team style assessment preferred.
• Ability to design and assess adversarial prompts within authorized scope.

Responsabilités

• Design authorized adversarial prompts to test AI safety boundaries.
• Create and assess scenarios in nuclear-security or nonproliferation domains.
• Review model reasoning for technical validity and evidence handling.
• Document judgments with explicit assumptions and reproducible checks.

Connaissances

Nuclear safeguards

Nonproliferation

AI evaluation

Red-team testing

Security analysis

Outils

Mass spectrometry

Radiochemical methods

Description du poste

Work you will do

• Design authorized adversarial prompts that test whether AI keeps an appropriate safety boundary, and explain why each response is acceptable or unsafe within the supplied professional context.
• Create and assess scenarios within your nuclear-security, safeguards, fuel-cycle, forensic or nonproliferation specialty.
• Review model reasoning for technical validity, evidentiary support and appropriate treatment of sensitive information.
• Document reference judgments and explicit evaluation criteria, identifying assumptions or missing evidence that affect the result.
• Use only authorized evaluation material and remain within the agreed analytical scope.

Required background and routes

• Professional nuclear-sector depth and experience with material control, safeguards or nonproliferation; select the route that matches your work.
• Nonproliferation: export classification/licensing/enforcement, end-user screening, NSG controls, NPT/CTBT or other treaty implementation, procurement/sanctions analysis or technical program assessment.
• Nuclear forensics: actual sample characterization and attribution, radiochemical or mass‑spectrometric methods, isotopic/age/impurity signatures, monitoring/detection or reference‑material methodology.
• Security engineering: material accountability, physical protection, vulnerability evaluation, insider‑risk/human-reliability programs or institutional security assistance.
• Fuel‑cycle engineering: practical conversion, fabrication, enrichment, reprocessing, spent‑fuel/waste systems, criticality controls, licensing or material‑flow modelling.
• Safeguards: IAEA, Euratom or SSAC/RSAC inspection practice, inventory/accountancy and MUF analysis, Additional Protocol implementation, NDA measurement, seals/surveillance or safeguards‑by‑design.

Preferred background

• Prior AI evaluation or red‑team practice, especially when supported by safeguards or nonproliferation work.

Deliverables

Submit the completed technical artifact or assessment with its supporting evidence, explicit assumptions, reproducible checks where applicable, and concise reasons for each material judgment. Address review findings within the agreed scope.

Location and schedule

Remote assignments are scheduled by agreement, with no guaranteed weekly volume. Availability planning can include 40 hours per week depending on the track. IXO confirms the applicable timing before work.

Pay and working terms

$70 - $80/task USD. Payment is for a completed task that meets the agreed acceptance criteria. The fee, task specification and revision requirements are agreed before work; an advertised time estimate does not determine payment. Applying does not guarantee an assignment. Use public, licensed or otherwise authorized material only; do not submit confidential employer information, personal data or restricted research. Security and safety testing stays within the expressly authorized evaluation environment.