Cyberlium

AI › Module 7 › Lesson 1

BeginnerModule 7Lesson 1/5

Secure SDLC AI

Secure SDLC for AI literacy — threat model, data governance, model card, release gate — on `$AI_LAB` dev workflow stubs.

15 min+40 XP3 quiz
Module progress1 of 5

Visual · t32_secure_sdlc_ai

Secure SDLC AI = named lifecycle rows. $AI_LAB. Original Cyberlium.

Opening

AI features need SDLC gates like any software — name secure AI lifecycle rows on YOUR lab workflow before shipping toy apps.

Secure SDLC for AI literacy names: AI threat modeling category (STRIDE-for-LLM literacy stub), training data provenance, model card / system card category, pre-release eval gate, monitoring and incident category, and change control for prompts/models. Analyst drafts SDLC checklist on `$AI_LAB` toy app repo — maps five lifecycle gates with owner stub — without skipping eval for 'demo speed,' without training on scraped PII, without deploying lab jailbreak findings to prod without RoE. Cyberlium teaches defender lifecycle vocabulary on YOUR repos — NOT shortcut to weaponized releases. Refused: skip eval gate, prod deploy without ticket, malware-gen in CI. Lab row: secure AI SDLC checklist ten items applied to toy app.

1. Named lifecycle gates

Threat model, data provenance, model card, eval gate, monitoring — five literacy anchors.

Each gate maps to artifact in YOUR `$AI_LAB` toy repo.

Command guide

Try these commands — Named lifecycle gates

═══ TOOLS & WEBSITES ═══ Browse / read these (authorized learning only — stay in YOUR lab / program scope)

NIST SSDF — https://csrc.nist.gov/projects/ssdf (secure development lifecycle) NIST AI RMF — https://www.nist.gov/itl/ai-risk-management-framework OWASP LLM Top 10 — https://owasp.org/www-project-top-10-for-large-language-model-applications/

═══ INSTALL ═══

Linux (Debian/Ubuntu):

Command — copy this

sudo apt install curl

macOS: Built-in

Windows: Built-in (PowerShell: Invoke-WebRequest)

═══ LINUX / macOS ═══

Command — copy this

export AI_LAB=${AI_LAB:-$HOME/cyberlium-lab/t32-ai}
curl -sS https://www.nist.gov/itl/ai-risk-management-framework | head -10
cat > "$AI_LAB/notes/secure-sdlc-ai-checklist.md" <<'EOF'
# Secure SDLC for AI — YOUR project notes
- [ ] Threat model: prompt injection, RAG poison, tool abuse
- [ ] Security requirements in user stories
- [ ] Pre-release red team on YOUR staging app (authorized)
- [ ] Monitoring: prompt/response logging with PII redaction
- [ ] Patch cadence for model + dependency updates
EOF

Command — copy this

grep '\[ \]' "$AI_LAB/notes/secure-sdlc-ai-checklist.md"

Primary tools to practice this lesson: grep, curl. Reference sites: NIST SSDF (https://csrc.nist.gov/projects/ssdf); NIST AI RMF (https://www.nist.gov/itl/ai-risk-management-framework); OWASP LLM Top 10 (https://owasp.org/www-project-top-10-for-large-language-model-applications/). Run every command in the box — install first, then the usage lines — only on YOUR lab / program scope.

2. Release discipline

No prompt/model change without eval rerun category — document on lab checklist.

Change control stub links Module 6 app security rows.

3. $AI_LAB boundary

YOUR toy repo or instructor stub only — not employer prod pipeline without RoE.

Refused: skip gates, scraped PII training, weaponized release artifacts.

4. What you ship: secure AI SDLC checklist

Ten lifecycle items applied to toy app + NEVER skip eval gate line.

5. What you record before the next lesson

Secure AI SDLC checklist path.

6. Wrong vs right: stranger SaaS vs YOUR toy LLM

Worked failure — same MSF word, opposite target. Right never needs a café Wi-Fi or classmate laptop.

  • Wrong

    Ship toy app to prod skipping eval because 'it's just AI demo.'

  • Right

    Secure AI SDLC checklist from `$AI_LAB` toy repo. Next: Eval and Guards.

Mission: secure AI SDLC checklist

1) Name five SDLC gates. 2) Ten-item checklist on toy app. 3) Owner stub per gate UTC. 4) Write NEVER skip eval gate line.

Stuck? Ask Cyberlium AI Mentor

Ask Mentor: “Model card — minimum literacy fields?”

Knowledge Check

1

APPLY: Secure SDLC AI uses:

Multiple choice

Knowledge Check

2

APPLY: True or False: Eval gate optional for AI demos.

True or False

Knowledge Check

3

APPLY: SDLC gates include:

Multiple choice

← Previous

Answer all 3 knowledge checks to continue. (0/3 answered)