SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
OpenAI's Preparedness team is seeking a Researcher focused on frontier cybersecurity risks to help design and implement an end-to-end mitigation stack that reduces severe cyber misuse across OpenAI's products. This role sits at the intersection of AI safety, security, and product engineering—addressing the urgent challenge that as AI models become increasingly capable agents, the same systems that accelerate productivity can also accelerate exploitation.
You will design and implement mitigation components spanning prevention, monitoring, detection, and enforcement for model-enabled cybersecurity misuse. Working closely with product and engineering teams, you'll integrate safeguards across product surfaces to ensure protections remain consistent, low-latency, and scale with usage and evolving model capabilities.
Key responsibilities include evaluating technical trade-offs (coverage, latency, model utility, privacy), collaborating with risk and threat modeling partners to align mitigation design with anticipated attacker behaviors, and executing rigorous testing and red-teaming workflows. You'll stress-test the mitigation stack against evolving threats—novel exploits, tool-use chains, automated attack workflows—across different product surfaces, then iterate based on findings.
This is urgent, fast-paced work with far-reaching implications for the company and society. You'll need demonstrated experience with deep learning and transformer models, proficiency in PyTorch or TensorFlow, strong fundamentals in data structures and algorithms, and experience with LLM training methods (distillation, supervised fine-tuning, policy optimization). Experience designing and deploying technical safeguards for abuse prevention at scale is essential. Background in cybersecurity is a plus. Above all, you should be passionate about AI safety and motivated to make cutting-edge AI models safer for real-world use.
