Principal AI Security Engineer

Principal AI Security Engineer

Role

This is a hands-on architecture role: deep research, secure systems design, and prototyping, applied across product lines with minimal direction and broad latitude to set the technical agenda. You will set platform-wide direction for securing the AI development lifecycle, define the security architecture, agents, and models operating within it, and represent ServiceNow's AI security point of view externally. You'll be a technical authority other engineers and other teams look to on emerging AI-specific threats, working in close partnership with the AI security research team - drawing on their findings and feeding real-world attack surfaces back into their agenda.

What you'll do

• Own the security strategy and architecture for our Agentic AI ecosystem, LLMs, and models across product lines
• Lead threat modeling for the most complex and novel GenAI/agentic surfaces: cross-agent autonomy, multi-agent tool orchestration, and emerging attack classes not yet standardised in the industry
• Set the architectural standard for securing model/RAG/data pipelines platform-wide, including access-control and data-boundary models that other teams are expected to build against
• Define what "secure-by-design" means for agentic systems at ServiceNow, and drive adoption across engineering orgs
• Represent ServiceNow's AI security posture externally (standards bodies, industry publications, conferences) and internally to executive stakeholders
• Mentor and technically guide engineers on AI-specific security work

Qualifications

To succeed in this role, you’ll need:

• Bachelor’s / Master's degree or PhD in Computer Science or a related technical field, specialisation in Security or AI/ML or an exceptional, well-evidenced track record substituting for it
• 10+ years of software engineering experience, including a track record of owning the architecture of complex systems at enterprise scale
• A track record of setting technical direction beyond your own team - architecture or standards other teams adopted, senior engineers you've levelled up, and comfort operating where the problem isn't yet defined
• 7+ years in security engineering - network and systems security, security protocols, design reviews, and threat modeling - with a focus on AI-specific work in recent years
• Multiple demonstrated engagements threat modeling and/or securing LLM, GenAI, or agentic systems, with evidence of platform-level or cross-team impact
• Personal ownership of prompt injection/jailbreak defense, guardrail architecture, or AI red-teaming programs - ideally having built the program/methodology, not just executed within one
• Deep, applied experience securing model, RAG, or data pipelines, including having designed the access-control and data-governance model others build against
• Fluency with the OWASP LLM Top 10 (or equivalent) to the point of extending, critiquing, or contributing to such frameworks - not just applying them
• Deep hands-on experience with agentic AI frameworks (e.g., LangChain, LangGraph)
• Clear communication, a collaborative default, and a bias toward learning fast in a field that changes constantly
• Bonus
• Strong command of auth protocols (OAuth 2.0, OIDC, PKCE, API Keys) and agentic protocols (MCP, A2A), including having designed identity/consent models for them
• Built and open-sourced (or productized) security tooling/automation for AI systems now used beyond one team
• Track record of CVEs, top-tier publications, or invited talks specifically in AI/ML or LLM security
• Prior experience setting AI security strategy at a platform or company level, not just a single product
• 5+ years of programming experience in Java or Python