Skip to content
Roadmap0/111
not startedstudy day 4 of 6 · covers original day 121

Design incident severity levels for data exposure

P1 AssistantP2 RAG platformP3 Agent systemS11M10C8
not started0/5 blocks
0/5
Jump & context
The five hoursRead todayResourcesEvidenceNotesBoth sources

Week 16: Cross-Project Red Teaming Completion & Operational Incident Response Drills

The five hours

Tap a block to start its 60/120/60/30/30 minute alarm.

  1. R31 — NIST AI RMF incident response

    R42 — GenAI incident categories

    Read the assigned sections; map the mechanism, threat model and trade-offs.

  2. Define formal AI Incident Severity Levels (SEV-1 to SEV-4) and classification criteria covering: cross-tenant PII/data exposure, unauthorized tool/agent side effects, model/supply-chain compromise, prompt-injection bypass, and denial-of-wallet exhaustion.

    Build, integrate, debug and improve the increment in the active project.

  3. No separate task text for this block on day 94 — the sourced work sits in the block above. Keep the time box.

    Test the abuse case, failure path, authorization bypass or resource limit.

  4. Incident severity classification matrix integrated into P1's C8 incident-diagnosis runbook.

    Add or run regression tests; record the 9-point evidence record.

  5. Answer the 8 Socratic questions without notes: What? Why? How? What if? Why not? How can I prove it? Where can it break? What have I learned?

    Re-explain the hardest concept aloud, inspect one edge case, update the gap log.

  6. total5 h

Read today · inside the concepts block

R31 R42

Read only the named sections needed for today's task. Reading sits inside the 60-minute concepts block; it is not extra time.

Free resources for this day

5 curated
  • docsNIST (NIST AI 600-1)
    Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile

    Provides the generative-AI risk categories (data privacy, information integrity, value chain) that your SEV classification criteria must explicitly cover.

    R42moved 2026-11-11
  • videoCodeLucky · 5:32
    SEV1 SEV2 SEV3 Explained: Incident Severity Levels Guide

    Lays out what distinguishes severity tiers and who gets paged at each, the scaffold for extending SEV-1 to SEV-4 with AI-specific triggers.

    R31checked 2026-11-11
  • videoMike Chapple · 11:54
    CertMike Explains Incident Response Process

    A clear walkthrough of the response lifecycle from a well-known security author, so each severity level maps to concrete preparation, detection and containment duties.

    R31checked 2026-11-11
  • bookGoogle SRE Book (O'Reilly, free online)
    Managing Incidents

    Shows the cost of unmanaged response versus declared incidents with roles and a live state document, which is what makes a cross-tenant exposure call escalate consistently.

    R31checked 2026-11-11
  • docsNIST
    AI Risk Management Framework | NIST

    Use the Manage function to bind each severity level to an owner, a response action and a documentation requirement so classification drives action rather than paperwork.

    R31checked 2026-11-11

Evidence record · 9-point standard

0/9

A lab is complete only when it produces evidence. Installing PyRIT, Garak, Presidio or TransformerLens is not completion. Saved on this device only.

Day notes

empty

The two source formulations

verbatim · nothing summarised

Project & Stage Target

Operational Security (All Projects, C8) | S11, M10

Core Implementation

Define formal AI Incident Severity Levels (SEV-1 to SEV-4) and classification criteria covering: cross-tenant PII/data exposure, unauthorized tool/agent side effects, model/supply-chain compromise, prompt-injection bypass, and denial-of-wallet exhaustion.

Verification & Evidence

Incident severity classification matrix integrated into P1's C8 incident-diagnosis runbook.

Read Today (60m Concepts Block)

R31 (NIST AI RMF incident response), R42 (GenAI incident categories).

---