Role Overview

The Researcher, Robustness & Safety Training position at OpenAI in San Francisco is a pivotal role within the Model Safety Research team. This team is dedicated to advancing AI safety capabilities, ensuring that OpenAI's most powerful models can be deployed safely to benefit society. As AI systems grow more capable, new challenges emerge in enforcing safety policies, robustness against adversaries, privacy protection, and trustworthiness in critical domains.

In this role, you'll conduct cutting-edge research on topics like Reinforcement Learning from Human Feedback (RLHF), adversarial training, and robustness. You'll implement these innovations directly into OpenAI's core training pipelines and launch safety features in products used by millions. This isn't just research—it's about shaping the future of safe AGI, collaborating across teams to meet the highest safety standards, and making a tangible impact on humanity.

OpenAI's commitment to safety is unwavering. The Safety Systems team leads efforts to deploy models responsibly, learning from real-world use while mitigating risks. If you have 4+ years in AI safety, a PhD in ML or related fields, and a passion for aligned AGI, this San Francisco-based role offers unparalleled opportunities to influence global AI safety standards.

Key focus areas include balancing safety with helpfulness, defending against malicious actors, securing user privacy, and building trust in high-stakes applications like healthcare or autonomous systems. Your work will directly contribute to OpenAI's charter for universally beneficial AI.

Key Responsibilities

As a Researcher in Robustness & Safety Training at OpenAI, your responsibilities will span research, implementation, and strategy:

• Lead state-of-the-art research in AI safety, specializing in RLHF, adversarial training, and model robustness techniques.
• Develop and implement novel methods within OpenAI’s core model training infrastructure for enhanced safety.
• Deploy safety improvements into production products, ensuring real-world reliability.
• Define research directions to make AI systems safer, more aligned, and resilient to adversarial threats.
• Collaborate with Trust & Safety, legal, policy, and engineering teams to uphold rigorous safety protocols.
• Perform comprehensive evaluations of model safety, pinpointing risks and crafting mitigation strategies.
• Engineer defenses against adversarial attacks, maintaining performance in hostile environments.
• Tackle privacy risks through advanced techniques like differential privacy in large models.
• Enhance fairness and reduce biases across diverse deployment scenarios.
• Build trustworthy AI for safety-critical domains, validating behaviors under uncertainty.
• Analyze deployment data to iteratively improve safety systems.
• Publish and present findings, advancing the broader AI safety field.
• Mentor junior researchers and contribute to team strategy.

These tasks position you at the intersection of research and deployment, ensuring OpenAI's AI benefits society safely.

Qualifications

To excel as a Researcher, Robustness & Safety Training in OpenAI's San Francisco office, you should possess:

• 4+ years in AI safety research, with expertise in RLHF, adversarial training, robustness, fairness, and biases.
• Ph.D. or equivalent in computer science, machine learning, statistics, or related disciplines.
• Hands-on experience with safety protocols for deploying large-scale AI models.
• Deep knowledge of deep learning architectures, optimization, and scaling laws.
• Strong programming skills in Python, PyTorch/TensorFlow, and ML frameworks.
• Demonstrated passion for OpenAI’s mission and alignment with its safety charter.
• Proven ability to collaborate in cross-functional, fast-paced environments.
• Track record of publishing in top ML conferences (NeurIPS, ICML, ICLR).
• Experience evaluating model vulnerabilities and proposing scalable fixes.
• Understanding of real-world AI risks like jailbreaks, data poisoning, and misuse.
• Excitement for tackling open problems in AGI safety.

Ideal candidates thrive in collaborative settings and are driven by the challenge of safe AI deployment.

Salary & Benefits

OpenAI offers competitive compensation for this senior Researcher role in San Francisco, estimated at $320,000 - $500,000 USD yearly, including base salary, equity, and bonuses. Total compensation reflects experience and impact.

Benefits include:

• Comprehensive medical, dental, vision coverage.
• 401(k) with generous matching.
• Unlimited vacation and flexible hours.
• Parental leave up to 20 weeks.
• Professional growth stipend ($10K+ annually).
• Onsite fitness center, catered meals, wellness programs.
• Relocation support for San Francisco move.
• Equity in OpenAI, a leader in AGI.
• Mental health resources and employee assistance.
• Commuter benefits and subsidized housing options.
• Access to cutting-edge hardware and compute clusters.
• Opportunities for global conferences and sabbaticals.

This package supports top talent focused on AI safety innovation.

Why Join OpenAI?

OpenAI is the pioneer in safe AGI development, with products like ChatGPT transforming the world. Joining the Safety Systems team in San Francisco means working with world-class researchers on problems that matter: making powerful AI safe for everyone.

Our culture emphasizes transparency, collaboration, and impact. You'll shape safety standards for future AI, publish influential work, and deploy changes affecting billions. San Francisco's vibrant tech ecosystem complements OpenAI's innovative environment.

With a mission to benefit humanity, OpenAI invests heavily in safety. This role offers intellectual challenge, massive scale, and the chance to define AGI safety. Past team members have advanced RLHF globally—your contributions could do the same.

OpenAI provides resources unmatched elsewhere: unlimited compute, expert peers, and direct mission alignment. If you're passionate about robust AI, this is your opportunity to lead.

How to Apply

Ready to advance AI safety at OpenAI? Submit your resume, cover letter highlighting AI safety experience, and links to publications or GitHub. Include why you're excited about OpenAI's mission.

Applications are reviewed on a rolling basis. Top candidates advance to research interviews, safety deep-dives, and team fits. We prioritize diverse perspectives aligned with our charter.

Apply now via OpenAI's careers page for the Researcher, Robustness & Safety Training role in San Francisco. Shape the future of safe AI today!