AI Security & Control Researcher

Visa sponsorship: Apollo Research says on this job page that it sponsors visas in the UK and US. It notes sponsorship is not guaranteed for every role or candidate.

Apollo Research tests frontier AI models before release and researches AI 'scheming'. It also runs Watcher, a security product that monitors coding agents. This researcher helps design threat models and control methods against AI agents that may be misaligned or compromised.

What you would do:

• Build threat models that treat coding agents like possible insider threats with access to code, credentials and networks
• Map out attack paths, in the style of frameworks such as MITRE ATT&CK, adapted for AI agents
• Define concrete, testable security levels for Watcher
• Help improve how well and how securely Watcher works

This is an individual contributor role at a start-up, with room to shape the job and possibly move into management later.

Apply directly on Apollo Research's website using the button on this page.