Open AI training and model evaluation roles that call for Adversarial Testing expertise, gathered from 4 platforms. Part of Data & Machine Learning.
Pay figures cover the 75 open jobs that publish an hourly rate in USD. They are published rates, not a guarantee of income. Last checked on October 8, 2026.
This contract role asks security-minded people to probe AI systems built on the OpenClaw platform by running red-teaming exercises and writing adversarial prompts. The work suits cybersecurity practitioners who enjoy finding weaknesses and who can document their findings clearly for technical and non-technical readers.
This remote contract role asks you to write original, inventive prompts that test how advanced AI models reason, respond, and handle edge cases. It suits people with strong writing skills and broad curiosity, and no technical or AI background is needed.
This freelance role asks you to write code and to evaluate code produced by AI models, helping make AI coding assistants more accurate and reliable. You will review AI outputs for correctness, efficiency, and security, then explain your judgments in clear written feedback. It suits experienced programmers who can work independently on a flexible, remote schedule.
This role involves evaluating how AI models reason about offensive security concepts and identifying flaws in their exploit chain reasoning. Penetration testers with hands-on red team experience will test model outputs across reconnaissance, exploitation, privilege escalation, and lateral movement, then write accurate attack paths that reflect real-world tradecraft when models fall short.
You will evaluate AI-generated code by running models on real engineering tasks, analyzing their output against production standards, and stress-testing them to find failures. This role is ideal for experienced software engineers who want to contribute to AI model improvement through rigorous code review and red-teaming.
You will work on evaluating frontier AI models to test their ability to distinguish legitimate chemical requests from malicious uses. As a chemical defense researcher, you will write targeted questions, evaluate model responses against a defined policy framework, and document your judgments with rigorous technical justifications. This role requires deep expertise in chemical defense to draw the line between routine professional questions and those with high potential for misuse.
Posted September 14, 2026
You will contribute to evaluating AI models as an expert in explosives and energetic materials. Your role involves writing sharp technical questions in your field, evaluating model responses against a safety policy, and documenting correct answers with technical justification. This work requires the ability to distinguish legitimate professional questions from suspicious requests that could serve malicious purposes.
Posted September 14, 2026
You will contribute to evaluating AI models for radiological safety by writing test requests at different difficulty levels and judging whether the model's responses meet safety standards. This role requires recognized expertise in radiological crisis management, capable of distinguishing legitimate professional questions from potentially dangerous requests.
Posted September 14, 2026
You will write and evaluate requests designed to test the ability of AI models to distinguish legitimate technical questions from dangerous requests in the field of energetic materials. This role requires deep expertise in post-incident analysis and regulatory inspection to draw the line between routine professional uses and potential misuse.
Posted September 14, 2026
You will contribute to the evaluation of frontier AI models as an expert in nuclear materials and non-proliferation measures. Your job is to write sophisticated test questions in your field and evaluate whether models correctly refuse dangerous requests while answering legitimate questions, drawing a subtle line between civil and military applications.
Posted September 14, 2026
You will join a team of chemistry and chemical safety experts tasked with testing the capabilities of frontier AI models to assess risks of misuse of legitimate purposes. Drawing on your expertise in forensic chemistry or toxicology, you will write sophisticated test questions, evaluate model responses against defined policy criteria, and develop reference answers with technical justifications.
Posted September 14, 2026
This role involves assessing the ability of AI models to correctly judge the potential for misuse of technical requests concerning energetic materials and propulsion. As a specialist, you will write progressively complex questions, evaluate model responses against a defined policy, and document your analyses to guide the behavior of AI systems when facing real-world risks.
Posted September 14, 2026
You will help test the robustness of AI models against malicious risks as a radiation protection specialist. Your role consists of designing deliberately ambiguous technical requests, evaluating the model's responses according to a defined security policy, and documenting best practices, while knowing how to distinguish legitimate professional questions from malicious intentions.
Posted September 14, 2026
You will evaluate how AI models handle sensitive requests in industrial hygiene and process safety, distinguishing legitimate professional questions from malicious bypass attempts. You will write graduated test requests, assess model responses against a defined policy standard, and document your judgments with technical reasoning accessible to non-specialists. This role is for experienced practitioners who have evaluated real exposures and process risks in professional settings.
Posted September 14, 2026
You will write and evaluate test cases to verify that an AI model knows how to refuse dangerous requests while answering legitimate questions in the field of nuclear nonproliferation. This role requires professional expertise in export controls, treaty implementation, or nuclear program analysis, so you can draw the line between routine compliance questions and circumvention attempts.
Posted September 14, 2026
You will contribute to evaluating frontier AI models as a nuclear fuel cycle expert. Your role will consist of writing nuanced technical questions and judging whether the model's answers comply with a safety policy, distinguishing between legitimate requests and potentially dangerous queries. This work requires a deep understanding of the parameters that differentiate a civilian application from a diversion for proliferation purposes.
Posted September 14, 2026
This job involves evaluating the capacity of AI models to distinguish legitimate requests from malicious uses in the field of nuclear medicine and medical isotopes. You will write nuanced test prompts and judge whether the model's responses comply with a security policy, while explaining your decisions in writing. This role requires specialized expertise to draw the line between routine professional questions and those concealing dangerous intentions.
Posted September 14, 2026
You will participate in evaluating cutting-edge AI models as a nuclear security specialist. You will write calibrated test prompts across three risk levels to judge whether the model correctly distinguishes legitimate professional questions from dangerous requests, then you will evaluate and document its responses according to a defined policy standard.
Posted September 14, 2026
You are an engineer specialized in propulsion and pyrotechnic systems. You will write calibrated technical questions to test an AI model's ability to distinguish legitimate requests from bypass attempts, then evaluate its responses and document the technical reasoning that justifies each judgment.
Posted September 14, 2026
Join a team of radiation safety experts to test the robustness of AI models against requests related to dual-use applications. You will write nuanced evaluation questions, assess model responses against strict safety criteria, and document your technical judgment to distinguish legitimate requests from dangerous ones.
Posted September 14, 2026
Add your resume once. Every open job is scored against your profile, and we email you only when a strong match appears. Free.
FAQ
76 Adversarial Testing AI training jobs are open on SideHustler today. Listings are rechecked against the official postings: the latest check on this list was on October 8, 2026.
Among the 75 open roles that publish an hourly rate in USD, pay runs from $15 to $150 per hour, with a median of $68. These are the rates the platforms publish, not a guarantee of income or hours.
Mercor (50), Alignerr (23), DataAnnotation (2) and Terac (1). You apply on the platform itself, which handles screening, contracts and payment.
The most requested areas of expertise on these listings are Adversarial Testing, Red Teaming, AI Evaluation, Vulnerability Assessment, AI Safety. Each listing states its own requirements.
Apply now opens the official posting on the platform, where you complete the application. You need no account on SideHustler. To see which roles fit your background first, add your resume and every open job is scored against your profile.
