SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
OpenAI's Preparedness team is hiring a Researcher to support preparations for accelerated AI development and recursive self-improvement. The role focuses on mitigating frontier risks from increasingly capable AI systems that can plan, execute, and adapt in the real world.
Key responsibilities include:
- Anticipating and addressing misalignment risks that may emerge as AI capabilities advance, particularly loss-of-control scenarios
- Designing and implementing pre-deployment risk assessments and control measures
- Developing scalable oversight practices that remain effective even as model capabilities exceed human performance
- Building automated auditing approaches to detect severe model misalignments in production traffic
- Conducting rigorous testing and red-teaming of measurements for model misbehavior (reward hacking, sandbagging, scheming)
- Designing experiments and evaluations to understand model alignment and safety-relevant capabilities
- Prototyping technical mechanisms for verifying compliance with AI safety agreements
- Tracking progress toward automation of technical work to inform alignment investments
- Identifying and addressing blindspots in mitigation strategies
The work alternates between hypothesis-driven research and turning insights into interventions that impact production models. You'll translate open-ended objectives into concrete research directions, build scrappy prototypes, iterate rapidly, and secure buy-in from other teams as needed.
Ideal candidates are exceptional technical executors with strong strategic and research taste—able to prioritize effectively in domains with weak feedback loops. Experience in ML research, AI alignment, AI verification, or related fields is valued. You should be passionate about mitigating recursive self-improvement risks and driven to do work that positively impacts the future of AI development.
