Radiologicals Safety for Redteaming

In short

This freelance remote mission asks you to test how AI models handle questions about radioactive material. You will check whether a model answers routine professional questions fully while refusing dangerous ones. It may suit you if you have truly been responsible for radioactive sources, through custody, licensing, security or emergency response, and can explain tricky decisions clearly.

What you will do

• Write tough single-turn questions in your field, sorted into benign, dual-use and adversarial groups.
• Check whether an AI model's answers follow a set safety policy.
• Write a reference answer that shows what a correct response looks like, with the technical reasons behind it.
• Explain each judgment in writing so that a non-specialist can follow your reasoning.

What you need to apply

• Real hands-on responsibility for radioactive material, not just work nearby, such as source custody, licensing, or security and response.
• The ability to tell a routine professional question from one that hides another goal.
• Strong written explanations. Earlier technical writing, published research or expert witness work is a plus, so include a sample or link.
• No classified, export-controlled or NDA-covered information in the work. If you have such obligations, say so in your application.

Who can apply

This is a freelance contract that is fully remote, and it is open to applicants anywhere in the world, with no country restriction stated.