CompTIA SecAI+ objectives and the drills that cover them

1.1

Compare and contrast various AI types and techniques used in cybersecurity: generative AI, ML, statistical learning, transformers, deep learning, GANs, NLP, LLMs, SLMs; model training (supervised, unsupervised, reinforcement, federated, fine-tuning, epochs, pruning, quantization); prompt engineering (system/user prompts, zero/one/multi-shot, templates)

• AI/ML Fundamentals for Security
  • Automation Bias — 96% Confidence Impossible Travel Alert Is Wrong
• Data Lineage, Integrity & RAG
  • AI Triage Review — RAG Context Window Truncation Caused Missed Evidence
• NLP & Deep Learning in Security
  • NL2KQL Validation — Catching a Schema Hallucination Before It Runs

1.2

Explain the importance of data security in relation to AI: data processing (cleansing, verification, lineage, integrity, provenance, augmentation, balancing); data types (structured, semi-structured, unstructured); watermarking; RAG (vector storage, embeddings)

• Data Lineage, Integrity & RAG
  • AI Triage Review — RAG Context Window Truncation Caused Missed Evidence
• NLP & Deep Learning in Security
  • NL2KQL Validation — Catching a Schema Hallucination Before It Runs

1.3

Explain the importance of security throughout the life cycle of AI: business use case alignment, data collection (trustworthiness, authenticity), preparation, model development/selection, evaluation, deployment, validation, monitoring/maintenance, feedback/iteration, human-centric design (human-in-the-loop, oversight, validation)

• AI-Driven Threat Landscape
  • AI Triage Review — AI Correctly Identifies Phishing but Misattributes the Affected User

2.1

Given a scenario, use AI threat-modeling resources: OWASP LLM Top 10, OWASP ML Security Top 10, MIT AI Risk Repository, MITRE ATLAS, CVE AI Working Group, threat-modeling frameworks

• AI Model Access Control & API Security
  • IAM Audit — Enterprise Chatbot with Excessive RAG Permissions
• AI Supply Chain Security
  • Data Lineage Audit — Trace AI Training Data and Inference-Time Access
• Execution Sandboxing for AI Agents
  • IAM Remediation — AI Agent Service Account Sprawl
• Model Poisoning & Data Integrity
  • Data Lineage Audit — Trace AI Training Data and Inference-Time Access
• Output Filtering & Guardrail Configuration
  • OPA/Rego Policy — Block AI Access to PHI/PCI-Tagged Data
• Prompt Injection Defense
  • AI Agent Output Validation — Retry Wrapper for Malformed JSON Responses

2.2

Given a scenario, implement security controls for AI systems: model controls (evaluation, guardrails, prompt templates); gateway controls (prompt firewalls, rate limits, token limits, input quotas, modality limits, endpoint access controls); guardrail testing and validation

• AI Model Access Control & API Security
  • IAM Audit — Enterprise Chatbot with Excessive RAG Permissions
• AI Deployment Security
  • IAM Hardening — Bedrock Model Invocation Open to All Principals
• Execution Sandboxing for AI Agents
  • IAM Remediation — AI Agent Service Account Sprawl

2.3

Given a scenario, implement appropriate access controls for AI systems: model access, data access, agent access, network/API access

• Adversarial Machine Learning Defense
  • Data Lineage Audit — Trace AI Training Data and Inference-Time Access
• Model Poisoning & Data Integrity
  • Data Lineage Audit — Trace AI Training Data and Inference-Time Access
• Prompt Injection Defense
  • AI Agent Output Validation — Retry Wrapper for Malformed JSON Responses

2.4

Given a scenario, implement data security controls for AI systems: encryption (in transit, at rest, in use); data safety (anonymization, classification labels, redaction, masking, minimization)

No drill mapped yet

2.5

Given a scenario, implement monitoring and auditing for AI systems: prompt monitoring (query, response), log monitoring/sanitization/protection, response confidence level, rate monitoring, AI cost monitoring (prompts, storage, response, processing), auditing for quality/compliance (hallucinations, accuracy, bias/fairness, access)

No drill mapped yet

2.6

Given a scenario, analyze evidence of an attack and suggest compensating controls: attacks (backdoor, trojan, prompt injection, poisoning, jailbreaking, input manipulation, bias introduction, guardrail circumvention, model inversion/theft, supply chain, transfer learning, model skewing, output integrity, membership inference, insecure output, DoS, sensitive disclosure, insecure plugin, excessive agency, overreliance); compensating controls (prompt firewalls, guardrails, access controls, data integrity, encryption, prompt templates, rate limiting, least privilege)

No drill mapped yet

3.1

Given a scenario, use AI-enabled tools to facilitate security tasks: tools (IDE plugins, browser plugins, CLI plugins, chatbots, personal assistants, MCP server); use cases (signature matching, code quality/linting, vulnerability analysis, automated pentesting, anomaly detection, pattern recognition, incident management, threat modeling, fraud detection, translation, summarization)

• AI-Augmented Threat Detection
  • AI Verdict Override — Authorized Backup Misidentified as Exfiltration
• NL2KQL & AI Query Validation
  • NL2KQL Validation — Catching a Schema Hallucination Before It Runs
• AI Verdict Validation & Automation Bias
  • Automation Bias — 96% Confidence Impossible Travel Alert Is Wrong

3.2

Explain how AI enables or enhances attack vectors: deepfakes (impersonation, misinformation, disinformation), adversarial networks, reconnaissance, social engineering, obfuscation, automated data correlation, automated attack generation (vector discovery, payloads, malware, honeypot, DDoS)

• AI-Driven Incident Response
  • AI Triage Review — AI Recommends "Monitor"; Analyst Determines Escalation Is Warranted
• NL2KQL & AI Query Validation
  • NL2KQL Validation — Catching a Schema Hallucination Before It Runs

3.3

Given a scenario, use AI to automate security tasks: scripting tools (low-code, no-code), document synthesis/summarization, incident response ticket management, change management (AI-assisted approvals, automated deployment/rollback), AI agents, CI/CD (code scanning, SCA, unit testing, regression testing, model testing, automated deployment/rollback)

• Behavioral Analytics & AI Monitoring
  • AI Triage Review — Unrelated Events Correlated by Shared NAT IP
• AI Verdict Validation & Automation Bias
  • Automation Bias — 96% Confidence Impossible Travel Alert Is Wrong

4.1

Explain organizational governance structures that support AI: AI Center of Excellence, AI policies/procedures; AI-related roles (data scientist, AI architect, ML engineer, platform engineer, MLOps engineer, AI security architect, AI governance engineer, AI risk analyst, AI auditor, data engineer)

• Global AI Regulations
  • OPA/Rego Policy — Require Audit Logging for All AI Agent Actions
• NIST AI Risk Management Framework
  • OPA/Rego Policy — AI Separation of Duties: Report Generator Cannot Approve

4.2

Explain risks associated with AI: responsible AI (fairness, reliability/safety, transparency, privacy/security, differential privacy, explainability, inclusiveness, accountability, consistency, awareness training); risks (bias, data leakage, reputational loss, accuracy/performance, IP risks, autonomous systems); shadow AI

• NIST AI Risk Management Framework
  • OPA/Rego Policy — AI Separation of Duties: Report Generator Cannot Approve

4.3

Summarize the impact of compliance on business use and development of AI: EU AI Act, OECD standards, ISO AI standards, NIST AI RMF, corporate policies (sanctioned vs unsanctioned, private vs public models, sensitive data governance), third-party compliance evaluations, data sovereignty

• Responsible AI & Ethical Frameworks
  • OPA/Rego Policy — Require Audit Logging for All AI Agent Actions