AI Safety & Oversight Researcher

Society for AI Security; San Francisco (CA); Hybrid; USD 150,000 - 240,000; Full time

Job summary

OpenAI in San Francisco, CA is seeking a Researcher in Agent Safety, Oversight and System Mitigations to advance safety for increasingly capable AI agents operating in real environments. You will design and evaluate system-level controls, collaborate with a Codex harness engineering team, and run red-team assessments to probe potential failures.

Candidates with strong systems or security instincts and a background in AI control or security are welcome.

Qualifications

• Background in AI safety or security with rigorous reasoning about agent behavior.
• Experience designing and evaluating mitigation mechanisms.
• Ability to build robust experimental infrastructure and evaluations.
• Interest in frontier AI alignment and safety.

Responsibilities

• Design, build, and evaluate system-level controls for agent actions.
• Collaborate with Codex harness engineering to productionize controls.
• Red-team end-to-end agentic systems to test for data exfiltration and harmful use.
• Measure safety-productivity tradeoffs and reduce latency.

Skills

Security research

AI safety

Agent safety

Systems thinking

Tools

Codex harness

Sandboxing

Process isolation

Job description

OpenAI in San Francisco, CA is seeking a Researcher in Agent Safety, Oversight and System Mitigations to advance safety for increasingly capable AI agents operating in real environments. You will design and evaluate system-level controls, collaborate with a Codex harness engineering team, and run red-team assessments to probe potential failures.

Candidates with strong systems or security instincts and a background in AI control or security are welcome.