Senior AI Security Researchers

Microsoft•Redmond, WA
•$119,800 - $304,200

About The Position

Microsoft Security Research conducts security research at the intersection of offense, defense, and intelligence. The organization works to protect AI systems from attacks as organizations adopt large language models, copilots, and autonomous agents. This role combines security research expertise with practical knowledge of modern AI and its vulnerabilities. The researcher will conduct realistic AI attack simulations and use findings to improve Microsoft security products and customer protections. The role involves shaping AI attack simulation methodology, building tooling, influencing product direction, and advancing both attacks against and defenses for AI systems.

Requirements

  • Doctorate in Statistics, Mathematics, Computer Science, Computer Security, or a related field; or a master's degree in one of these fields plus 3 or more years of relevant experience; or a bachelor's degree in one of these fields plus 4 or more years of relevant experience; or equivalent experience.
  • Relevant experience includes software development lifecycle, large-scale computing, threat analysis or modeling, cybersecurity, vulnerability research, and/or anomaly detection.
  • Ability to meet Microsoft, customer, and/or government security screening requirements, including passing a Microsoft background check and Microsoft Cloud background check upon hire or transfer and every two years thereafter.

Nice To Haves

  • Doctorate in a listed field plus 3 or more years of relevant experience; or a master's degree plus 6 or more years; or a bachelor's degree plus 8 or more years; or equivalent experience.
  • 6 or more years of experience in security research and/or AI/ML security.
  • Security experience finding and exploiting real vulnerabilities, such as penetration testing, red teaming, vulnerability research, or application security, with an adversarial mindset.
  • Practical understanding of modern AI, including LLMs, prompting, retrieval-augmented generation, fine-tuning, and agentic or tool-using systems.
  • Hands-on familiarity with agentic AI systems, red-teaming harnesses, and AI tooling, including agents, tool-using LLMs, orchestration and evaluation frameworks, LLM APIs, and attack harnesses.
  • Hands-on experience attacking AI systems, including prompt injection, jailbreaks, guardrail bypass, data exfiltration, or agent abuse, in research or professional settings.
  • Python skills for building attack automation, evaluation harnesses, and research tooling.
  • Familiarity with AI security and safety frameworks such as MITRE ATLAS, the OWASP Top 10 for LLMs, and Responsible AI principles.
  • Ability to translate offensive findings into practical defenses and communicate clearly with engineers and executives.
  • Contributions to AI security research, publications, CTFs, or open source.

Responsibilities

  • Lead the design and execution of realistic AI attack simulations, using findings to improve security products and customer protection.
  • Develop and execute direct prompt injection and jailbreak attacks that bypass system prompts, safety instructions, and alignment.
  • Conduct indirect or cross-domain prompt injection attacks through untrusted documents, web pages, emails, tickets, code, and retrieval-augmented generation sources.
  • Conduct multi-turn, crescendo, and social-engineering-style attacks that steer models toward unsafe behavior.
  • Test multimodal injection through images, audio, and files carrying hidden instructions.
  • Test guardrails, content filters, and safety systems for bypasses, including through encoding, obfuscation, and adversarial phrasing.
  • Investigate sensitive-data exfiltration and leakage, including disclosure of system prompts and grounding data, and cross-tenant or cross-session leakage.
  • Test agentic and tool-use abuse, including excessive agency, confused-deputy attacks, unsafe autonomous actions, and manipulation of tools, plugins, and connected systems.
  • Investigate data and model attacks, including training-data and RAG poisoning, model extraction, model inversion, membership inference, and adversarial examples or evasion.
  • Build scalable AI red-teaming tooling and automation, including attack harnesses and evaluation frameworks for generating attack variations, running large evaluations, and continuously testing AI systems.
  • Partner with product, engineering, Responsible AI, and detection teams to translate findings into mitigations such as improved prompts, filters, grounding controls, agent guardrails, and detections.
  • Evaluate AI defenses, including detections, safety classifiers, filters, and monitoring, and recommend ways to close security gaps.
  • Research emerging AI attacker techniques and map them to frameworks such as MITRE ATLAS and the OWASP Top 10 for LLM Applications.
  • Deliver executive-level briefings, technical reports, and prioritized, actionable recommendations.
  • Shape AI simulation methodology, mentor team members, and drive long-term innovation in Security for AI.
© 2026 Teal Labs, Inc
Privacy PolicyTerms of Service