Data Scientist, Preparedness

Details

The Preparedness team is part of OpenAI's Safety Systems organization and works to identify, track, and prepare for catastrophic risks related to frontier AI models. The team focuses on capability assessment, evaluations, internal red teaming, mitigation systems, and coordination on AGI preparedness.

The role involves building, evaluating, and continuously improving mitigations that prevent extreme harms from AI systems. The successful candidate will independently structure ambiguous problems, conduct rigorous analyses, and translate findings into actionable product and policy changes. The role includes developing mitigation intelligence and monitoring systems to detect issues early, measure effectiveness over time, and reduce both over-blocking and under-blocking.

Responsibilities

• Evaluate and improve mitigation systems, including classifiers and detection pipelines across domains such as biosecurity, cybersecurity, and emerging risk areas.
• Diagnose false positives and false negatives through deep error analysis, root-cause investigation, and clear recommendations for mitigation adjustments.
• Build monitoring and measurement frameworks to track mitigation effectiveness over time and across user segments and use cases.
• Identify trends in over-blocking versus under-blocking, quantify customer impact, and propose prioritized interventions.
• Develop insights from customer feedback, complaints, and usage patterns to detect shifts in adversarial behavior and system failure modes.
• Expand risk monitoring into new areas, including cybersecurity threats and model loss-of-control or sabotage scenarios, in partnership with domain experts.
• Communicate results to technical and executive stakeholders through concise narratives, decision-ready metrics, and clear tradeoffs.

Requirements

• Significant experience in data science or applied analytics in high-stakes domains such as security, trust and safety, abuse prevention, fraud, platform integrity, or reliability.
• Strong foundations in experimentation, causal thinking, and/or observational inference, with the ability to design robust measurement under imperfect data.
• Fluency in SQL and Python, or equivalent tools, for analysis, modeling, and building monitoring workflows.
• Experience building metrics, dashboards, and operational monitoring that meaningfully changes outcomes.
• A track record of driving cross-functional impact with engineering, product, and research partners.
• Strong preference for cybersecurity data science experience, including exposure to threat modeling, adversarial dynamics, abuse patterns, or security telemetry.
• Experience with classifier evaluation, calibration, thresholding, and error analysis at scale.
• Familiarity with detection systems in adversarial settings, including evasion, distribution shift, and feedback loops.
• Trust and safety experience is helpful but not required.
• Genuine interest in AI safety, alignment, and catastrophic risk prevention.
• Ability to operate autonomously, structure analyses end-to-end, communicate clearly to executives, and turn analysis into product changes.

Benefits

• Base salary of $265,000–$380,000 per year, plus equity.
• Medical, dental, and vision insurance, with employer contributions to Health Savings Accounts.
• Pre-tax accounts for health, dependent care, and commuter expenses.
• 401(k) retirement plan with employer match.
• Paid parental, medical, and caregiver leave.
• Paid time off, paid company holidays, office closures, and paid sick or safe time as required by law.
• Mental health and wellness support.
• Employer-paid basic life and disability coverage.
• Annual learning and development stipend.
• Daily office meals and eligible meal delivery credits.
• Relocation support for eligible employees.
• Additional benefits may include charitable donation matching and wellness stipends.

OpenAI is an equal opportunity employer and provides reasonable accommodations to applicants with disabilities. Background checks are administered in accordance with applicable law.