Top AI Ethics Jobs in San Francisco Bay Area, CA

The Senior AI Risk Advisor will lead AI risk assessments, develop governance frameworks, evaluate AI vendors, and engage with stakeholders on AI adoption.

The CTO will lead Fathom's technical strategy, focusing on AI governance, building partnerships, defining standards, and advancing AI safety practices.

The Manager will oversee a team focused on technical governance and AI safety, managing projects and stakeholders while contributing to research.

The AI Quality Associate evaluates AI outputs for quality, develops benchmarks and rubrics, validates workflows, and collaborates with finance experts to ensure high-quality AI-assisted outputs.

As a Research Engineer, you will tackle AI safety challenges, develop evaluation methodologies, conduct adversarial testing, and work on model alignment and interpretability. You'll bridge theoretical and practical applications of AI safety research and collaborate with product teams to implement safety solutions.

The role involves implementing AI safety strategies, building evaluation frameworks, and developing safeguards against scientific risks from AI models. Candidates must possess strong technical knowledge and communicate complex concepts effectively.

Convert interpretability research into production product features: design and build full-stack interfaces, APIs, workflows, and integrations for model evaluation, debugging, monitoring, and deployment. Partner with researchers, engineers, designers, and customers to deliver reliable, performant, and user-friendly tooling for understanding model internals.

As a Technical Product Intern, you'll prototype user safety features, conduct competitive analysis, and translate compliance frameworks into product logic, focused on real-time AI enforcement.

The role involves evaluating AI-generated content, providing feedback, and ensuring cultural relevance and quality in various media types as a contractor.

Review text prompts and AI-generated videos to identify mismatches, visual or behavioral errors, and tag incorrect prompt text. Add concise annotations, perform self-QA, participate in calibration, document recurring model failures, and meet throughput and quality targets using structured annotation platforms.