# Researcher, Recursive Self-Improvement Safety

**Company:** [OpenAI](https://hotfix.jobs/companies/openai)
**Location:** San Francisco, CA
**Role:** AI Research
**Salary:** $295k – $445k/yr
**Skills:** Machine Learning, ai alignment, ai verification, scalable oversight, automated auditing, red teaming, model monitoring, misalignment detection, reward hacking, chain-of-thought monitoring, model evaluations, ai safety, Risk Assessment, control systems, training interventions
**Posted:** 2026-08-07

> Conduct strategic, technically rigorous research to anticipate and mitigate loss-of-control risks from increasingly capable AI systems, including recursive self-improvement. The role combines hypothesis-driven research, rapid prototyping, safety evaluations, monitoring, and institutionalizing effective interventions.

## Job Description

## Responsibilities
- Anticipate future risks from increasingly capable AI systems, including risks associated with recursive self-improvement.
- Translate open-ended safety objectives into concrete, prioritized research directions.
- Build rapid prototypes and iteratively improve them into established components of safety pipelines.
- Develop pre-deployment risk assessments, control measures, training interventions, and institutional practices for loss-of-control risks.
- Establish scalable oversight and model-misbehavior monitoring practices for highly capable models.
- Develop automated auditing approaches to identify severe misalignment in production traffic and elicit tail risks before deployment.
- Test and red-team measurements of reward hacking, sandbagging, scheming, and other loss-of-control behaviors.
- Design experiments and evaluations to study model misalignment and safety-relevant capabilities.
- Develop model organisms of misbehavior and training interventions that improve safety-relevant capabilities.
- Prototype technical mechanisms for verifying compliance with potential future AI safety agreements.
- Measure progress toward automation of technical staff to inform alignment and security investments.
- Identify and address blind spots in recursive self-improvement safety cases.
- Secure organizational buy-in and communicate technical work clearly.
- Collaborate with or manage other staff as needed.

## Requirements
- Exceptional technical execution ability.
- Strong strategic and research judgment in domains with weak feedback loops.
- Passion for mitigating risks associated with recursive self-improvement.
- Willingness to prioritize work based on its potential positive impact on the future of AI development.

## Nice-to-haves
- Experience in machine learning research, AI alignment, AI verification, or a related domain.

## Compensation
- Annual salary: **$295,000–$445,000**

## Similar roles

- [Researcher, Frontier Risk Mitigations](https://hotfix.jobs/jobs/7ff5dad3-4092-4016-af83-7697cc0a93a8) - OpenAI - San Francisco, CA - $295k – $445k/yr
- [Research Engineer / Research Scientist / AI Systems Engineer, RSI](https://hotfix.jobs/jobs/9de7d179-a863-41e6-afd6-3f2c7ab08ad0) - OpenAI - San Francisco, CA - $295k – $445k/yr
- [Research Engineer / Research Scientist - Personal AGI, Personalization](https://hotfix.jobs/jobs/8af0d507-7137-40c8-98e1-11be16336041) - OpenAI - San Francisco, CA - $295k – $555k/yr
- [Researcher, Multimodal Safety](https://hotfix.jobs/jobs/2cc84f9d-4c40-42d1-858a-f8dcb305ef5f) - OpenAI - San Francisco, CA - $295k – $445k/yr
- [Researcher, Synthetic RL](https://hotfix.jobs/jobs/1d48e4f7-9589-45b3-96ed-05276e32344e) - OpenAI - San Francisco, CA - $295k – $445k/yr

**Apply:** https://hotfix.jobs/jobs/1f8fd130-4760-4d3a-af28-47681ec30870
**Canonical:** https://hotfix.jobs/jobs/1f8fd130-4760-4d3a-af28-47681ec30870