Frontier AI Risks Lead

Official role information

ABOUT THE TEAM The Intelligence and Investigations team seeks to rapidly identify and mitigate abuse and strategic risks to ensure a safe online ecosystem in close collaboration with our internal and external partners. Our efforts contribute to OpenAI's overarching goal of developing AI that benefits humanity. This role focuses specifically on AI Safety: understanding and mitigating risks created or amplified by increasingly capable AI systems. It is not a cybersecurity, information security, or corporate security role. The Strategic Intelligence & Analysis (SIA) team provides safety intelligence for OpenAI’s products by monitoring, analyzing, and forecasting real-world abuse, geopolitical risks, and strategic threats. Our work informs AI safety mitigations, product decisions, and partnerships, ensuring OpenAI’s tools are deployed responsibly across critical sectors. ABOUT THE ROLE We are looking for a Frontier AI Risks Lead to help us understand potential harms and misuse of AI in a time of rapid, sustained change. We seek to understand how developments in AI could intersect with misuse and abuse, accelerating existing harm areas and creating novel risks. We seek to scan available signals and use strategic foresight methodologies to enable proactive detection and mitigation of frontier AI risks. This is an AI safety role focused on frontier and systemic risks, including model misalignment, recursive self-improvement (RSI), multi-agent interaction, loss of control, runaway agents, and related emerging failure modes. In this role, you will help provide a strategic-level perspective on a range of frontier AI safety areas, producing actionable understanding of issues relevant to OpenAI’s platforms, systems, and broader mission. Utilizing mixed quantitative and qualitative methodologies, you will spot early warning signs, pull threads on potentially concerning behavior, and turn weak signals into clear, prioritized risk calls. You will focus on upstream ecosystem scanning, competitive benchmarking, and external narrative and risk sense-making. Your work will inform cross-functional partners across research, product, policy, preparedness, and safety to guide mitigations that keep increasingly capable AI systems beneficial and reduce the likelihood and impact of misuse or other emerging harms. IN THIS ROLE, YOU WILL - Map and prioritize emerging risks at the frontier of AI - Build and continuously refine a clear picture of emerging signals and trends that could affect the AI ecosystem through upstream and external scanning. - Design and maintain harm taxonomies that provide foresight and warning about how AI harms and misuse may manifest over the next 0-24 months and beyond. - Contribute to an evergreen frontier risk register and prioritization framework that surfaces the top issues by severity, prevalence, exposure, and trajectory. - Detect and deep dive into emerging abuse patterns - Create comprehensive approaches to horizon scanning, competitive benchmarking, and external narrative/risk sense-making. - Stay current on abuse trends ranging from state actor misuse to criminal activity, drawing from the work of internal organizational and cross-functional partners. - Connect individual incidents into system-level stories about actors, incentives, product design weaknesses, and cross-product spillover—whenever possible spotting these incidents or even hypothesizing them before they hit our surfaces. - 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. - Work with Global Affairs and Communications teams to share findings in ways that reinforce OpenAI’s role as a leader in AI safety. - 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 fast-evolving AI environments are safe, including model misalignment, interpretability, recursive self-improvement, multi-agent interactions, loss-of-control scenarios, runaway agents, and societal or systemic risks. - Work with data science and visualization colleagues to shape monitoring views and dashboards that highlight leading indicators and unusual changes from signals spotted off platform to determine whether these are manifesting in user behavior or abuse patterns. - Pioneer new uses of our own technologies to scale detection and transform workflows. - Provide strategic analysis and future-looking perspectives - Produce concise but comprehensive strategic intelligence estimates that provide full context about a given interest area that includes confidence levels based on observed data to inform judgments and recommendations. - Run scenario analyses that explore how frontier AI risks might evolve over the next 6–24 months, including how multi-agent misalignment, recursive self-improvement, loss of control, or runaway-agent behavior could manifest in important economic sectors. - Help design and run tabletop exercises for internal and partner audiences that distill manifest and latent risks at the frontier of AI and identify mitigations. - Benchmark OpenAI’s risk profile and mitigations against external incidents and other platforms, highlighting gaps, strengths, and opportunities. - Shape safety readiness for new products - Contribute to product readiness and launch reviews by laying out expected abuse modes based on broad, upstream understanding. - 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.…

Review the linked official posting for the complete current wording, application questions and any location-specific terms.

Verified at source level during this run from OpenAI's credential-free public Ashby job-board API. The provider returned this record in its current published collection. Missing salary, deadline, visa, funding, eligibility and location facts were not inferred.