Chemicals Safety for Redteaming
- Contract: Contract
- Who can apply: Open worldwide
- Posted Sep 9, 2026
In short
Mercor is building a team of chemistry experts to test AI models on chemistry questions. Your job is to tell safe, useful answers apart from risky ones, so the AI helps with normal questions and refuses truly dangerous ones. If you have hands-on lab experience and enjoy careful written reasoning, this freelance role may suit you.
What you will do
- Write hard single-turn prompts in chemistry and label each one as benign, dual-use, or adversarial.
- Check whether an AI model's answer to each prompt was handled correctly, using a set policy standard.
- Write a reference answer, what a good response looks like, and the technical reasoning behind it.
- Explain each judgment in clear writing that a non-specialist can follow.
What you need to apply
- Hands-on experience in synthesis, analytical chemistry, or chemical safety.
- Strong writing skills; past technical writing, published research, or expert witness work is a strong plus, and you should include a sample or link.
- Experience with AI evaluation or red-teaming is preferred but not required.
- Willingness to read and write about misuse scenarios for long periods, knowing you can pause or step away at any time.
Who can apply
The posting does not limit applicants by country, so people anywhere in the world can apply for this remote freelance contract, which has 899 openings. If you hold classified, export-controlled, or NDA-bound knowledge, you can still apply and explain this in your application.
My tip
We suggest sending a writing sample or a link to your published work with your application, since the posting says this is a strong signal.