Model Policy Manager, Multimodal Safety

Highlights

• Shape safety for multimodal models like GPT-Live and ChatGPT Images.
• Design model policies for audio, image, video, and omni-modal behavior.
• Drive safety across training, evaluation, red-teaming, and deployment.

What they offer

• Equity
• Relocation assistance

Ideal candidate

A pragmatic, evidence-driven safety thinker with strong judgment on multimodal AI risks who thrives turning ambiguity into measurable policy and collaborating in fast-moving research teams.

Requirements

• Strong judgment about real-world risks of advanced multimodal AI systems
• Experience turning ambiguous safety questions into data-driven policies and evaluation criteria
• Treat policy as measurable end-to-end system and diagnose gaps across data, graders, and safeguards
• Strong technical judgment to design policies that can be trained, measured, and supervised at scale
• Technical fluency using AI tools to accelerate policy development and analyze failures
• Comfortable hands-on with model data and evaluation results, inspecting examples and failure patterns
• Thrive in fast-paced, collaborative research environments amid shifting priorities
• Pragmatic, evidence-driven approach to reducing risk while preserving beneficial uses
• Hands-on experience driving consensus and action in ambiguous spaces