AI Social Risk Analyst

OpenAI•San Francisco, CA

10d

About The Position

We are looking for an AI social risk analyst to sit at the frontline of a new category of challenges: AI-enabled social experiences. As AI systems like Sora and advanced chat models become part of messaging, group conversations, creator tools, and branded experiences, new forms of challenges could emerge in human–AI–human interactions. In this role, you will own the analytical view of safety and abuse risks in AI-social environments: Sora content and sharing, group chats, messaging, and AI-assisted brand and creator experiences. You will spot early warning signs, pull threads on potentially concerning behavior, and turn weak signals into clear, prioritized risk calls. Your work will guide mitigations that keep users, brands, and communities safe while allowing productive, creative uses of these tools to thrive.

Requirements

Significant experience (typically 5+ years) in trust and safety, integrity, security, policy analysis, or intelligence work focused on social media, messaging, online communities, or adjacent environments.
Demonstrated ability to analyze complex online harms (e.g., harassment, coordinated abuse, scams, synthetic media, influence operations, brand safety issues) and convert analysis into concrete, prioritized recommendations.
Strong analytical skills and comfort working with both qualitative and quantitative inputs, including: (1) Casework, incident reports, OSINT, product context, and policy frameworks. (2) Basic metrics and trends in partnership with data science (e.g., harm prevalence, severity profiles, exposure, escalation rates).
Strong adversarial and product intuition, able to foresee how actors might adapt AI-social and creative tools for misuse and evaluate how product mechanics, incentives, and UX decisions influence risk.
Experience designing and using risk frameworks and taxonomies (e.g., harm classification schemes, severity/likelihood matrices, prioritization models) to structure ambiguous spaces and support decision-making.
Proven ability to work cross-functionally with product, engineering, data science, operations, legal, and policy teams, including pushing for clarity on tradeoffs and following through on mitigation work.
Excellent written and verbal communication skills, including experience producing concise, executive-ready briefs and explaining sensitive, complex issues in grounded, concrete terms.
Comfort operating in fast-changing, ambiguous environments: you can identify weak signals, form hypotheses, test them quickly, and adjust as the product and threat landscape evolves.

Responsibilities

Map and prioritize the AI-social risk landscape
Build and continuously refine a clear picture of how AI is used in social-like products (e.g., Sora-powered clips, group chats, messaging assistants, creator tools).
Design and maintain harm taxonomies tailored to AI-mediated communication (e.g., synthetic harassment, coordinated AI-assisted brigading, synthetic identity/brand misuse, reputational and intimate harms).
Maintain a risk register and prioritization framework that surfaces the top issues by severity, prevalence, exposure, and trajectory.
Detect and deep dive into emerging abuse patterns
Partner with investigations, operations, and product teams to surface new patterns of misuse across Sora, chats, and partner integrations.
Run structured deep dives on incidents, from synthetic impersonation and scams to targeted harassment or coordinated influence using AI-generated media.
Connect individual incidents into system-level stories about actors, incentives, product design weaknesses, and cross-product spillover.
Turn analysis into actionable risk intelligence
Translate findings into clear, ranked risk lists and concrete proposals for mitigations that product, safety, and policy teams can execute on.
Collaborate with Safety Systems, Integrity, and Product to scope solutions such as classification improvements, UX guardrails, friction, enforcement flows, and detection signals.
Track whether mitigation work is landing: follow key indicators, pressure-test assumptions, and push for course corrections when the data demands it.
Build early warning and measurement capabilities
Help define the core metrics and signals that indicate whether AI-social environments are safe (e.g., key harm prevalence, severity distributions, escalation rates, brand safety issues).
Work with data science and visualization colleagues to shape monitoring views and dashboards that highlight leading indicators and unusual changes in user behavior or abuse patterns.
Propose targeted probes, structured reviews, and experiments that surface new risk modes around major launches and feature changes.
Provide strategic analysis and future-looking perspectives
Produce concise, decision-ready briefs on AI-social risks for leadership, safety forums, and partner teams.
Run scenario analyses that explore how AI-social harms might evolve over the next 6–24 months (e.g., how attackers might adapt to Sora, how group chats could be used for coordination, likely pressure points for brands and public figures).
Benchmark OpenAI’s AI-social risk profile and mitigations against external incidents and other platforms, highlighting gaps, strengths, and opportunities.
Shape safety readiness for social-like AI products
Contribute to product readiness and launch reviews by laying out expected abuse modes, risk tradeoffs, and monitoring/response plans.
Turn risk insights into practical guidance for internal teams (product, marketing, partnerships, comms) and, where appropriate, external partners using OpenAI technologies in social and brand contexts.
Develop reusable frameworks, playbooks, FAQs, and briefing materials that make it easier for the broader organization to understand AI-social risks and respond consistently.