AI Security & LLM Security — Skills Plan
Planned skill map for AI Security & LLM Security. The curriculum will be developed and released in stages.
Threat-model AI applications and practice controls for prompt injection, retrieved data, model outputs, tool permissions, privacy, monitoring, and incident response.
Potential roles: AI Security Engineer · Application Security Engineer
Before you start
- Networking and web security
- Python and API fundamentals
- Cybersecurity foundations
- LLM application architecture
What you’ll build
- Threat-model a RAG assistant
- Test a tool-using workflow
- Create an AI incident-response playbook
Prerequisites
PLANNED- Networking and web security
Build the Networking and web security knowledge needed to follow the AI Security & LLM Security foundations.
- Python and API fundamentals
Build the Python and API fundamentals knowledge needed to follow the AI Security & LLM Security foundations.
- Cybersecurity foundations
Build the Cybersecurity foundations knowledge needed to follow the AI Security & LLM Security foundations.
- LLM application architecture
Build the LLM application architecture knowledge needed to follow the AI Security & LLM Security foundations.
Foundations
PLANNED- AI threat modeling
Explain how AI threat modeling works and where it fits in the AI Security & LLM Security learning path.
- Trust boundaries and data flows
Explain how Trust boundaries and data flows works and where it fits in the AI Security & LLM Security learning path.
- Prompt injection concepts
Explain how Prompt injection concepts works and where it fits in the AI Security & LLM Security learning path.
- Sensitive-data handling
Explain how Sensitive-data handling works and where it fits in the AI Security & LLM Security learning path.
Core Skills
PLANNED- RAG permission enforcement
Use RAG permission enforcement to implement a working capability and verify its behavior in AI Security & LLM Security.
- Tool authorization and least privilege
Use Tool authorization and least privilege to implement a working capability and verify its behavior in AI Security & LLM Security.
- Output validation and safe rendering
Use Output validation and safe rendering to implement a working capability and verify its behavior in AI Security & LLM Security.
- Secrets and provider configuration
Use Secrets and provider configuration to implement a working capability and verify its behavior in AI Security & LLM Security.
- Data poisoning and supply-chain risks
Use Data poisoning and supply-chain risks to implement a working capability and verify its behavior in AI Security & LLM Security.
Advanced / Production
PLANNED- Adversarial testing and red teaming
Evaluate Adversarial testing and red teaming trade-offs against the performance, reliability, and safety needs of AI Security & LLM Security systems.
- Abuse monitoring and rate controls
Evaluate Abuse monitoring and rate controls trade-offs against the performance, reliability, and safety needs of AI Security & LLM Security systems.
- Model and dependency governance
Evaluate Model and dependency governance trade-offs against the performance, reliability, and safety needs of AI Security & LLM Security systems.
- Incident response for AI systems
Evaluate Incident response for AI systems trade-offs against the performance, reliability, and safety needs of AI Security & LLM Security systems.
- Privacy and retention review
Evaluate Privacy and retention review trade-offs against the performance, reliability, and safety needs of AI Security & LLM Security systems.
- Human oversight for high-impact actions
Evaluate Human oversight for high-impact actions trade-offs against the performance, reliability, and safety needs of AI Security & LLM Security systems.
Real-World Projects
PLANNED- Threat-model a RAG assistant
Deliver a working outcome with Threat-model a RAG assistant, then test it against the project requirements.
- Test a tool-using workflow
Deliver a working outcome with Test a tool-using workflow, then test it against the project requirements.
- Create an AI incident-response playbook
Deliver a working outcome with Create an AI incident-response playbook, then test it against the project requirements.
Career / Interview Preparation
PLANNED- AI security design review
Explain your approach to AI security design review, defend key trade-offs, and discuss what you learned.
- Threat and mitigation mapping
Explain your approach to Threat and mitigation mapping, defend key trade-offs, and discuss what you learned.
- Abuse-case analysis
Explain your approach to Abuse-case analysis, defend key trade-offs, and discuss what you learned.
- Security portfolio walkthrough
Explain your approach to Security portfolio walkthrough, defend key trade-offs, and discuss what you learned.
Discussion
0Have a better approach? Found something outdated? Share it — your knowledge helps everyone learning here.