AI Security Engineer
Open a video or resource link, watch it, then check its box to track your progress and unlock what comes next.
- Build agent threat models + adversarial testing foundations
- Enforce identity/tool boundaries + isolate untrusted execution
- Operate secure agents with SDL, supply-chain controls + SecOps telemetry
Benchmark: OpenAI Agent Security + Anthropic Product Sandboxing/Application Security (current Sep 2026) Market refresh: agent threat models, red teaming, identity/tool boundaries, isolation, secure SDLC and agent SecOps.
Who’s Hiring Right Now
US openings posted in the last 7 days at Anthropic, OpenAI — this role’s own benchmark employers — refreshed daily, as of 2026-11-01. Pulled directly from each company’s public job board; not exhaustive, and postings close fast — worth confirming the details on the employer’s own site before applying.
Enrichment
Build agent threat models + adversarial testing foundations
Agentic AI threat modeling + attack surfaces
Coverage: Security Fundamentals & Threat Modeling / Agentic Security
Prompt injection, jailbreaks + adversarial testing
Coverage: AI Red Teaming & Adversarial Testing
Enforce identity/tool boundaries + isolate untrusted execution
Agent identity, least privilege + MCP/tool boundaries
Coverage: Identity & Access Management (IAM) / Agentic Security / MCPs
Sandboxing, containers + runtime isolation
Coverage: Agentic Security / Cloud Security / Containers
Operate secure agents with SDL, supply-chain controls + SecOps telemetry
Secure SDLC, supply chain + policy enforcement
Coverage: DevSecOps / Cloud Security / AI & Agent Governance
Agent telemetry, detection + incident response
Coverage: Detection, Incident Response & Security Operations
Role-Readiness Proof / Capstone
Threat-model and harden a production agent: add per-agent identity and least privilege, tool/MCP authorization, sandboxed code execution, cloud/container controls, automated red-team tests, telemetry and real-time policy checks; detect a simulated compromise, investigate it, contain it and produce an incident plus secure-design review package.
Proof Rubric
- Demonstrate exploit-to-control evidence: threat-model coverage
- automated abuse tests
- identity/tool boundary enforcement
- isolation
- secure-SDLC controls
- high-signal telemetry/detections
- containment/incident response
- quantify residual risk and usability/performance tradeoffs