AI risk governance and incident leadership

Prerequisites: distinguish output from evidence; review the AI workflow reference. Public self-study is not official agency training, accreditation or a credential.

What you will be able to do

This lesson develops the ability to exercise risk ownership, escalation, and incident decisions. The setting is NORTHSTAR Mission Portfolio, a fictional public-service portfolio containing ten proposed AI and non-AI interventions, three synthetic vendors, constrained staff capacity, and a staged incident. Every person, organization, system record, metric, and outcome in the case is fictional. Do not replace the case with operational, classified, CUI, export-controlled, proprietary, employer-confidential, or third-party personal information.

By the end of the lesson, you will be able to:

• Demonstrate how to exercise risk ownership, escalation, and incident decisions.
• Apply that model to the synthetic NORTHSTAR Mission Portfolio environment.
• Produce a reviewable artifact that separates evidence, assumption, uncertainty, and authority.
• Recognize the critical failure and select a safe remediation path.

The performance standard is not confident language. It is a traceable decision, an artifact another qualified reviewer can inspect, a clear boundary around the learner’s authority, and a safe response to uncertainty. Knowledge checks are formative and do not award a certification or endorsement.

This curriculum runs self-study on VetsLearn. Where the text calls for an instructor, peer or formal review, use the stated acceptance criteria for self-review. Reflection questions support reflection; they do not replace independent assessment. A peer review is optional unless your actual organization requires it.

Part 1 Incident facts and recovery gates

Separate containment, preservation, official reporting, communication and recovery. Distinguish observed facts from hypotheses. Use designated security, privacy, records and mission routes; do not assert exposure, cause or legal conclusions without evidence. Recovery needs tests for each failure: denied unauthorized access, source validation, restoration and configuration regression. Urgency does not validate missing controls. Governance needs thresholds, owners and power to act, not just a risk register.

Run the four-inject tabletop

Time	Observed fictional fact	Protective action	Recovery evidence
09:00	6/20 wrong dates	Pause date guidance; manual contact	Source-support regression
09:10	Cross-role record appears	Restrict affected retrieval	Direct and semantic denial tests
09:20	Deadline pressure to reopen	Explain separate failures; retain fallback	Designated owner decision
09:40	Rollback fixes dates only	Keep access restriction	Denial and trace review still required

Communication draft: “We observed incorrect date outputs and a synthetic cross-role retrieval failure. The affected guidance and retrieval functions are restricted while responders assess cause and scope. Manual handling remains available. Reopening requires source checks and access-denial tests plus the responsible decision.” This distinguishes known facts from an unconfirmed breach hypothesis.

Recovery branch: if direct lookup is denied but paraphrased retrieval still returns the record, the access defect remains. If dates are corrected but sources are stale, currency evidence remains open. Preserve incident evidence according to authorized instructions; a classroom exercise does not direct handling of actual sensitive logs.

Feedback: a good after-action item names a mechanism and test, such as enforce role filters before model context and test semantic lookup. “Be more careful” is not a control repair.

Part 2 Synthetic worked example

Inject 1: wrong dates rise to six of twenty. Inject 2: a synthetic restricted record appears to another role. Inject 3: leadership demands reopening. Inject 4: rollback fixes dates but access remains broken. Keep affected access paused with a manual fallback. Fixing dates does not demonstrate access recovery.

Guided practice: identify the unsupported conclusion, name missing evidence and rewrite its scope. Change one assumption and identify which decision reopens.

Part 3 Feedback and repair

Pass when recovery covers both failures, uncertainty stays explicit and evidence is not destroyed outside the authorized process. One improved metric cannot reopen every function.

Retain the failed version, explain the criterion, correct the artifact and test a different invented case.

Next step: carry the artifact and failed cases into the next module lesson or capstone. Retain limitations when explaining your work.