Adversarial Testing AI training jobs

Safeguards Analyst / IAEA or Euratom Inspector

• Science & Research
• Platform: Mercor
• Location: United States

Mercor is looking for nuclear and safeguards experts to test the robustness of AI models against dual-use requests. You will write targeted prompts on the knife's edge between legitimate professional questions and dangerous requests, then evaluate the model's responses against a defined policy and write reference answers. This role is for practitioners with concrete experience in nuclear material verification and detecting diversion attempts.

Posted September 14, 2026

Source Security Specialist

• Health & Medicine
• Platform: Mercor
• Location: United States

You will contribute to evaluating frontier AI models as a source security specialist. Your role is to write prompts that test the model's ability to distinguish legitimate professional questions from potentially dangerous requests, then evaluate its responses against a defined policy standard. This work requires practical expertise in Category 1 and 2 radioactive source security, as well as strong technical writing skills to justify your assessments.

Posted September 14, 2026

Specialty Materials Chemist - Pyrotechnics or Propulsion

• Science & Research
• Platform: Mercor
• Location: United States

You will contribute to evaluating AI models as a specialist chemist in energetic materials, by writing technical requests that test the model's ability to distinguish legitimate questions from dangerous requests. You will need to master the fine line between civilian applications and misuse to design realistic scenarios and evaluate the model's responses.

Posted September 14, 2026

Synthetic / Process Chemist

• Science & Research
• Platform: Mercor
• Location: United States

You will write questions and scenarios to test an AI model's ability to correctly judge the potential for misuse of technical requests in chemistry, distinguishing legitimate questions from dangerous requests. This role requires practical expertise as a synthetic chemist or process engineer with experience scaling up reactions, and the ability to write clear technical justifications for non-specialists.

Posted September 14, 2026

Chemicals Safety for Redteaming

• Science & Research
• Platform: Mercor
• Location: United States

You are a chemistry expert with practical experience in synthesis, analytical chemistry, or chemical safety. You participate in evaluating AI models by writing calibrated test prompts across three risk levels, then assessing model responses against an established safety framework. This role requires both advanced technical expertise and the ability to document your judgments in an accessible way.

Posted September 9, 2026

Energetic Materials Expert for Redteaming

• Science & Research
• Platform: Mercor
• Location: United States

This role involves evaluating an AI model's responses to requests involving energetic materials and explosives, distinguishing between legitimate questions and dangerous requests. You will write progressively more complex test prompts, assess the model's responses against a defined safety policy, and provide written justification for your technical judgments. This role is intended for experts in materials chemistry specializing in propulsion and initiation systems, with solid practical experience in the field.

Posted September 9, 2026

Radiologicals Safety for Redteaming

• Health & Medicine
• Platform: Mercor
• Location: United States

Mercor is recruiting radiological safety experts to test the robustness of AI models against dual-use requests. You will write progressively complex prompts, evaluate model responses according to a defined policy, and document reference responses with technical justifications. This role requires deep expertise capable of distinguishing legitimate professional questions from potentially dangerous requests.

Posted September 9, 2026

Bilingual Croatian Generalist Expert - AI Safety

• Business, Consulting & Operations
• Platform: Mercor
• Location: United States
• Language: Croatian

This role involves evaluating and improving the safety of AI models by analyzing their behavior on sensitive topics in Croatian. You will write expert prompts, classify content according to structured guidelines, and identify adversarial formulations, as an English-Croatian bilingual speaker with no prior AI experience required.

Posted September 4, 2026

Bilingual Dutch (Belgium) Generalist Expert - AI Safety

• Languages, Translation & Voice
• Platform: Mercor
• Location: Belgium
• Language: Dutch

This role asks bilingual English and Dutch speakers to test how AI models respond to sensitive topics in Dutch as used in Belgium. The work combines language fluency with cultural judgment, and no prior AI experience is needed because the workflow is taught on the job.

Posted September 4, 2026

Bilingual Finnish Generalist Expert - AI Safety

• Languages, Translation & Voice
• Platform: Mercor
• Location: United States
• Language: Finnish

This role involves evaluating and improving the safety of advanced AI models in Finnish as a bilingual English-Finnish speaker. You will apply your language mastery and judgment to examine how these models handle sensitive topics, without requiring prior experience in AI or machine learning.

Posted September 4, 2026

Bilingual French Generalist Expert - AI Safety

• Languages, Translation & Voice
• Platform: Mercor
• Location: United States
• Language: French

This role consists of evaluating the safety of AI models in French by examining how they handle sensitive topics. You will write expert prompts, apply classification guidelines, and identify adversarial formulations, leveraging your bilingual mastery of French and English.

Posted September 4, 2026

Bilingual Norwegian Generalist Expert - AI Safety

• Data & Machine Learning
• Platform: Mercor
• Location: United States
• Language: Norwegian

This role consists of evaluating and strengthening the safety of advanced AI models by analyzing their behavior on sensitive topics in Norwegian. You will use your bilingual proficiency (English-Norwegian) and cultural judgment to identify gaps and bypass attempts, with no prior AI experience required.

Posted September 4, 2026

Bilingual Turkish Generalist Expert - AI Safety

• Data & Machine Learning
• Platform: Mercor
• Location: United States
• Language: Turkish

You will participate in improving AI model safety by evaluating how they handle sensitive subjects in Turkish. Your linguistic and cultural judgments will help identify and strengthen weaknesses in these systems when facing delicate content. No prior AI experience is required.

Posted September 4, 2026

Professionals: Designing Challenging AI Prompts

• Data & Machine Learning
• Platform: Terac
• Location: United States

This is a paid one-hour trial for people who know a professional or personal workflow well. Participants turn that workflow into a demanding prompt, test it in ChatGPT to find where the model fails, and make the prompt harder when the model succeeds. The work suits domain experts and power users who can judge AI outputs quickly and who are open to ongoing evaluation work.

Posted August 25, 2026

LLM Research Scientist (Pre-training & Computer Vision & Adversarial Robustness)

• Data & Machine Learning
• Platform: Mercor
• Location: United States
• Level: Intermediate

We are looking for experienced machine learning researchers to work on empirical research problems covering computer vision, language models, and robustness to adversarial attacks. You will train and improve end-to-end models, optimizing trade-offs between performance, efficiency, and robustness in resource-constrained contexts.

Posted August 5, 2026

AI Safety Experts - English & Danish

• Data & Machine Learning
• Platform: Mercor
• Location: United States
• Language: Danish

You join a red teaming team specialized in adversarial evaluation of conversational AI models. Your role consists of testing AI systems by exploring their vulnerabilities (jailbreaks, prompt injections, biases), generating high-quality data documenting these vulnerabilities, and producing reproducible reports to strengthen model safety. This position is for bilingual English-Danish experts with prior experience in red teaming or related fields (cybersecurity, adversarial ML, socio-technical analysis).

Posted July 30, 2026

AI Safety Experts - English & Dutch

• Data & Machine Learning
• Platform: Mercor
• Location: United States
• Language: Dutch

This role consists of testing the robustness and security of conversational AI models by subjecting them to sophisticated adversarial attacks. You will generate high-quality training data by identifying vulnerabilities, biases, and systemic risks that automated tests do not detect. The ideal profile has prior experience in red teaming, structured adversarial thinking, and the ability to clearly communicate security risks.

Posted July 30, 2026

AI Safety Experts - English & Finnish

• Data & Machine Learning
• Platform: Mercor
• Location: United States
• Language: Finnish

This role involves testing the limits and vulnerabilities of conversational AI models using adversarial techniques to identify security risks. You generate high-quality data that enables clients to improve the robustness and reliability of their AI systems. The ideal profile masters red teaming, structured adversarial thinking, and the ability to reproducibly document discovered flaws.

Posted July 30, 2026

AI Safety Experts - English & Indonesian

• Data & Machine Learning
• Platform: Mercor
• Location: United States
• Language: Indonesian

You will join an adversarial testing team to evaluate the robustness and safety of conversational AI models. Your job is to identify vulnerabilities by generating malicious inputs, annotating failures, and documenting systemic risks to enable clients to improve their AI systems. This role suits experts with experience in security testing and a natural ability to explore the limits of systems.

Posted July 30, 2026

AI Safety Experts - English & Malay

• Data & Machine Learning
• Platform: Mercor
• Location: United States
• Language: Malay

This role involves training conversational AI models by testing them adversarially to identify vulnerabilities and improve their safety. You will generate high-quality training data by annotating failures, classifying risks, and documenting reproducible attack cases. The ideal profile has prior experience in red teaming or cybersecurity, and masters a structured and methodical approach to testing the limits of AI systems.

Posted July 30, 2026