AI Security Engineer (GRC – Healthcare) | W2 Contract | USC Only |

The Job

The AI Security Engineer (GRC) serves as the organization's dedicated subject matter expert at the intersection of artificial intelligence and cybersecurity within a regulated healthcare environment. This role is responsible for evaluating AI vendors and technologies, establishing and enforcing secure AI implementation standards, and providing hands-on guidance to development and engineering teams adopting AI platforms such as Microsoft Copilot Studio, Azure AI Foundry, Snowflake Cortex, Claude Code, and other large language model (LLM)-powered tooling. 

Operating within the HIPAA-regulated landscape, this analyst will ensure AI integrations — including Model Context Protocol (MCP) servers, agentic workflows, command-line interfaces (CLIs), APIs, and third-party AI extensions — are architected and deployed in a manner consistent with NIST AI RMF, HITRUST, and organizational security policies. The role acts as a trusted advisor, security gatekeeper, and enabler for responsible AI adoption across the enterprise.

You Will

1.  AI Vendor & Technology Evaluation 

• Lead structured security assessments of AI vendors, platforms, and tools prior to organizational adoption or renewal 
• Evaluate vendor data handling practices, model training transparency and data residency 
• Assess the security posture of AI platforms including: 
• Microsoft Copilot Studio — plugin trust boundaries, connector authentication, Power Platform DLP policies 
• Azure AI Foundry — model deployment pipelines, private endpoint configuration, managed identity usage 
• Snowflake Cortex — data access controls in AI-generated SQL, Snowpark security, role-based privilege enforcement, Cortex function access policies, and query result exposure risks 
• Claude Code & Anthropic APIs — system prompt injection risks, tool use / agentic permissions, data retention settings 
• GitHub Copilot, Cursor, and other AI-assisted development tools — code telemetry and secret leakage exposure 
• Produce written Vendor Security Assessment Reports (VSARs) including risk ratings, compensating controls, and recommendations 
• Maintain an AI technology registry with risk classifications and review cadence schedules 

2.  Secure AI Implementation Guidance for Development Teams 

• Serve as the embedded security advisor to software engineering, data science, and clinical informatics teams adopting AI tooling 
• Define and enforce secure-by-default configurations for AI development environments and agentic systems 
• Review and approve MCP server configurations, ensuring: 
• Tool definitions follow least-privilege principles — no excessive file system, network, or shell access 
• Server authentication uses OAuth 2.0 / mTLS and does not rely on static API keys stored in plaintext 
• Transport layer security (TLS 1.2+) is enforced on all MCP server communications 
• Prompt injection attack surfaces are identified and mitigated in tool descriptions and system prompts 
• Logging and audit trails are enabled for all MCP tool invocations touching PHI or sensitive data 
• Establish CLI security standards for AI-assisted development tools (Claude Code CLI, GitHub Copilot CLI, Azure Developer CLI), including credential hygiene, shell history scrubbing, and token scope minimization 
• Conduct secure code review for AI integration code — with focus on prompt injection, insecure deserialization, and unsafe agentic action chains 
• Develop and maintain a library of reference architectures, secure configuration templates, and implementation checklists for approved AI platforms 

3.  AI Risk Management & Compliance 

• Maintain the organization's AI Risk Register aligned with NIST AI RMF (Govern, Map, Measure, Manage) 
• Ensure AI deployments comply with HIPAA Security Rule (45 CFR §164), HITECH Act obligations, and applicable state privacy laws 
• Conduct AI-specific Threat Modeling (STRIDE / PASTA) and red-team exercises targeting: 
• Prompt injection and jailbreak scenarios 
• Indirect prompt injection via external data sources (email, documents, web retrieval) 
• Model inversion and membership inference attacks on fine-tuned healthcare models 
• Data exfiltration through agentic tool chains 
• Track emerging AI threats and threat actor TTPs relevant to healthcare AI systems via MITRE ATLAS and sector ISACs 
• Participate in AI governance committee meetings and contribute AI security perspectives to organizational AI policies 

4.  Security Integration Reviews 

• Review AI integration architectures for network segmentation, data flow, and trust boundary enforcement 
• Validate that PHI is never transmitted to external AI models without de-identification or explicit BAA coverage 
• Assess retrieval-augmented generation (RAG) architectures for unauthorized data access and embedding extraction risks 
• Evaluate agentic AI workflows and multi-agent orchestration systems for privilege escalation and uncontrolled action chains 
• Provide security sign-off on AI infrastructure as part of the Change Advisory Board (CAB) process 

5.  Training, Awareness & Policy 

• Develop AI security training curricula for developers, data engineers, clinical staff, and IT personnel 
• Author and maintain AI security policies including: Acceptable Use of Generative AI, AI Vendor Onboarding Standards, MCP and Agentic System Security Policy, and Sensitive Data Handling in AI Contexts 
• Publish internal guidance and threat intelligence briefings tailored to clinical and technical audiences 

Your Qualifications

• Bachelor’s degree in Cybersecurity, Computer Science, Information Systems, or a closely related field 
• Master’s degree preferred; equivalent professional experience considered
• 7+ years of progressive experience in information security, with a minimum of 2 years focused on AI/ML security or applied AI technology evaluation 
• Demonstrated hands-on experience with one or more of the following: Copilot Studio, Azure AI Foundry, Claude / Anthropic APIs, OpenAI API, GitHub Copilot, or LLM agentic frameworks (LangChain, AutoGen, Semantic Kernel) 
• Experience working in a HIPAA-regulated environment; healthcare industry background strongly preferred 
• Proven track record conducting vendor risk assessments and producing executive-level risk documentation
• Deep understanding of LLM attack surface: prompt injection, indirect prompt injection, system prompt extraction, and model manipulation 
• Familiarity with AI red-teaming methodologies and tools (Garak, PyRIT, PromptBench) 
• Knowledge of OWASP Top 10 for LLM Applications 
• Understanding of AI model lifecycle risks: training data poisoning, supply chain risks in model registries (Hugging Face, Azure Model Catalog)
• Ability to audit and secure Model Context Protocol (MCP) server implementations including: 
• Reviewing tool definitions and permissions for least-privilege violations 
• Validating authentication mechanisms (no hardcoded credentials, proper token scoping) 
• Assessing stdio vs. SSE transport security implications 
• Identifying SSRF and command injection risks in custom MCP tool implementations 
• Experience securing AI CLIs including credential storage, environment variable exposure, and shell integration risks 
• Knowledge of agentic permission models — understanding when AI agents should require human-in-the-loop approval 
• Ability to evaluate multi-step AI workflow chains for unintended capability escalation
• Microsoft Copilot Studio: Plugin manifest security review, connector authentication, sensitivity label enforcement 
• Azure AI Foundry: Managed identity configuration, private endpoints, content filtering policy management, model deployment governance 
• Snowflake Cortex: Securing AI-generated SQL and Cortex LLM functions, Snowpark container security, column-level data masking, network policy enforcement, and OAuth integration for service accounts 
• Claude Code: System prompt construction, tool-use permission hardening, CLI credential isolation, API key scoping 
• GitHub Copilot Enterprise: Telemetry settings, suggestion filtering for secrets, IDE extension trust policies
• Strong grounding in identity and access management — OAuth 2.0, OIDC, SAML, managed identities, workload identity federation 
• API security: authentication schemes, rate limiting, input validation, and output sanitization for AI endpoints 
• Network security: micro-segmentation, private endpoints, WAF configuration for AI service ingress 
• SIEM/SOAR integration for AI audit log ingestion, anomaly detection, and automated response 
• Threat modeling methodologies: STRIDE, PASTA, and application of MITRE ATT&CK and ATLAS frameworks
• Thorough understanding of HIPAA Security Rule requirements and how they apply to AI data processing pipelines 
• Experience with HITRUST CSF controls relevant to AI and cloud-based processing of ePHI 
• Practical knowledge of NIST AI Risk Management Framework (AI RMF) — Govern, Map, Measure, Manage functions 
• Familiarity with EU AI Act classifications and their implications for healthcare AI systems (high-risk AI designation) 
• Experience reviewing BAAs and DPAs for AI vendor engagements