Threat Modeler Lead: OpenAI (Safety Systems Team)

The role owns OpenAI's holistic approach to anticipating future harms from powerful AI. The central function is to act as the "central node" connecting the technical findings from AI evaluations to the governance and policy decisions of the company.

Primary Risk Domains (Threat Models)
The role requires developing and maintaining threat models across two primary categories, reflecting the dual-use nature of advanced AI:

Misuse Risks (External Actors):

Bio/Chemical: Modeling how AI could lower the barrier for malicious actors to create biological or chemical threats (e.g., assisting in wet lab planning feasibility or operational planning).

Cyber: Modeling AI-enabled offensive cybersecurity capabilities (e.g., using the model to create novel, high-value exploits or attacks on critical infrastructure).

Attack Planning: Modeling the use of AI for complex, long-horizon, harmful planning.

Alignment Risks (Internal Control):

Loss of Control: Modeling risks where the AI system may pursue unintended goals, known as the "specification problem."

Self-Improvement: Modeling the risks of the AI autonomously replicating, adapting, or resisting shutdown.

Key Responsibilities
Develop Comprehensive Threat Models: Systematically anticipate how various threat actors might leverage frontier AI to achieve catastrophic outcomes.

Risk Forecasting: Combine technical foresight, adversarial simulation, and emerging trends to forecast when specific high-level capabilities might be reached.

Evaluation and Safeguards: Pair closely with technical partners to ensure capability evaluations map to and adequately cover the full spectrum of severe risks.

Mitigation Design: Translate threat models into actionable mitigation designs and size the remaining risk of deployed safeguards.

Policy & Governance Rationale: Serve as the thought partner and explainer for high-investment mitigation efforts, justifying the "why" and "when" of prioritization to technical, governance, and policy stakeholders.

Required Expertise
Candidates must bring deep experience in:

Threat Modeling, Risk Analysis, or Adversarial Thinking (adapted from security or national security domains).

Understanding of risks from frontier AI systems and a strong grasp of AI alignment literature.

Familiarity with how AI evaluations ("evals") work and how to connect their results to capability testing and safeguard sufficiency.

The ability to think in systems and naturally anticipate second-order and cascading risks.