Technical Program Manager - AI Safety Enforcement
MicrosoftRedmond, WA,US
Full-timePosted 2 days agoStill open, checked 2h ago
About the job
Analyze user prompts, model outputs, account activity, and other available evidence to assess respective risk, policy applicability, and appropriate enforcement or mitigation outcomes using AI Safety guidance and Responsible AI frameworks. Grade the user prompts/model outputs according to the policy. Apply investigative expertise across text, code, image, audio, video, and multimodal abuse scenarios Analyze available data to connect abuse signals, scope potential impact, and support evidence-based enforcement and mitigation decisions using existing processes and tools. Document abusive behaviors, indicators, tactics, techniques, and patterns involving generative AI to support repeatable investigations and improved detection. Develop, refine, and validate AI Safety taxonomies, investigative guidelines, classifier definitions, reviewer instructions, escalation criteria, and evaluation datasets. Translate investigation findings into actionable recommendations for detections, classifiers, model safeguards, enforcement strategies, policies, and product mitigations. Lead or contribute to priority AI misuse investigations, incidents, high-risk submissions, customer appeals, and complex escalations involving potential AI abuse. Bachelor's Degree AND 2+ years experience in engineering, product/technical program management, data analysis, or product development OR equivalent experience. 1+ year(s) of experience managing cross-functional and/or cross-team projects. These requirements include, but are not limited to the following specialized security screenings: Bachelor's degree in applicable field (either Cybersecurity, Chemical or Bio, Computer Science, Information Security, Data Analytics, Business, or a related field, OR 2+ years of experience in cybersecurity, security operations, threat intelligence, incident response, trust & safety, fraud, risk analysis, abuse investigations, or related operations, or equivalent professional experience). Experience conducting evidence-based security, CBRN abuse, or threat investigations, assessing available signals, and making decisions using structured policies, investigative frameworks, or enforcement guidelines. Ability to handle sensitive or potentially disturbing content while following company policies. Understanding of generative AI and large language model capabilities, limitations, and failure modes, with an emphasis on potential cyber or bio misuse and attempts to circumvent safeguards. Familiarity with cyber threat frameworks, taxonomies, indicators of compromise, and threat-intelligence platforms or standards such as MITRE ATT&CK, MITRE ATLAS, OWASP. Written and verbal communication skills, including the ability to document investigative evidence and explain cyber risks, findings, and mitigation recommendations to technical and non-technical audiences. Experience reviewing or investigating text, image, audio, video, or multimodal content using structured policies, taxonomies, or evaluation frameworks. Experience partnering with security, threat intelligence, engineering, data science, policy, legal, compliance, or customer-support teams on sensitive cyber investigations or escalations.
