AI Security Engineer

Role : AI Security Engineer

Location : Oakland, CA

Position Overview

The AI Engineer will help monitor and respond to how AI tools, agents, and connectors are used across the organization. Working with the Infrastructure Applications Support Manager within Digital Innovation and Technology (DigIT), and in close partnership with Security, this role will design and implement monitoring, alerting, and response capabilities using Microsoft Defender and other enterprise security tools. Scope includes enterprise AI tools such as Microsoft 365 Copilot, ChatGPT EDU, and Box AI, as well as AI agents, third-party connectors, and Model Context Protocol (MCP) connections. This is a remote, six-month contract engagement subject to extension should the demand exist.

Key Responsibilities

AI Usage Visibility and Discovery

Assess current security tools and configuration to identify AI monitoring gaps and what is needed to close them.

Configure cloud access security broker (CASB) and related capabilities to discover and classify sanctioned and unsanctioned AI applications, model providers, and MCP servers.

Establish usage baselines for approved enterprise AI tools to identify unusual activity.

Use data loss prevention and data security capabilities to identify sensitive data flowing into AI tools.

Agents, Connectors, and MCP Connections

Inventory AI agents, OAuth-connected apps, connectors, and MCP connections that touch data, including owner, permissions, and approval status.

Develop policies and alerts for high-risk patterns such as over-permissioned connectors, unapproved third-party AI apps, and personal AI accounts connecting to data.

Monitor AI agents for risky configurations, data access, and behavior, in line with the Agentic AI Guardrails.

Detection, Alerting, and Response

Develop detections, alerts, and hunting queries for AI-related risks such as data exfiltration, prompt injection, and anomalous agent activity.

Integrate audit and usage logs from enterprise AI platforms, where feasible, to centralize monitoring.

Partner with Security on triage and response runbooks defining severity, escalation, and actions per alert type.

Design monitoring that respects privacy requirements, including the Electronic Communications Policy, by favoring risk signals and metadata over content review.

Request Research, Risk Documentation, and Reporting

Research incoming requests for new AI tools, agents, connectors, and MCP servers, assess security risks, and recommend approval conditions.

Document risks, controls, and recommended practices aligned with the AI Acceptable Use Standard, the Agentic AI Guardrails, and UC policy IS-3.

Report on AI usage, risk trends, and open issues for security, program, and leadership audiences.

Minimum Qualifications

Bachelor's degree in Information Security, Computer Science, or a related field, or equivalent experience.

5 or more years of security engineering, security operations, or cloud security experience in an enterprise environment.

Hands-on experience with CASB, SIEM, and data loss prevention tools, including app discovery, policies, and alerts.

Experience writing detection and hunting queries, and scripting (for example, Python or PowerShell) for automation, log analysis, and API integrations.

Working knowledge of identity and access management, including OAuth consent, application permissions, and service accounts.

Understanding of AI security risks such as data leakage, prompt injection, over-permissioned agents and connectors, and shadow AI.

Strong written communication skills, including documenting risks, controls, and runbooks for varied audiences.

Ability to work independently and collaboratively in a remote, cross-functional environment.

Preferred Qualifications

Experience with Microsoft security tools such as Microsoft Defender and Microsoft Purview.

Experience securing AI agents and integrations, such as MCP servers, connectors, or AI APIs.

Familiarity with admin, audit, and compliance features of Microsoft 365 Copilot, ChatGPT EDU, or Box AI.

Familiarity with AI risk frameworks such as the NIST AI RMF or the OWASP Top 10 for LLM Applications.

Security certifications such as CISSP, CCSP, or vendor security certifications.

Familiarity with higher education or the University of California, including UC IS-3 and the Electronic Communications Policy.

Success Measures

Percentage of AI apps, agents, connectors, and MCP connections inventoried with owner, permissions, and approval status.

Priority AI risk scenarios covered by live, tuned detections.

Time to triage and resolve AI-related alerts.

Unapproved AI apps and high-risk connectors remediated or approved.

Turnaround time for dispositioning AI security requests.

Monitoring operated by staff after hand-off.