The formerly missing introductory worked teaching and guided decision-record practice are supplied. This status means introductory material is available; independently marked learner performance remains unresolved.
Original editorial teaching and fictional application, informed by the linked primary passages. Compare your reasoning with the model; this is not independently marked learner work, an official examination key or a replacement for the selected course.
What this guide contributes.
Primary route: ISC2 CC. These are selected related scope positions, not a claim that the guide teaches the entire objective/application/statement. Full scope and remaining depth are in the coverage register.
See this guide’s exact scope links and remaining depth
What the source review checked
- Worked adoption decision; asset/use/data/affected people/manual alternative
- Risk appetite, tolerance, approval/stop authority
- Accountability and reviewed decision artifact
- Data/model protection and privacy
- Ethical harms and control trade-offs
- Plausible harms, selected controls, residual risk and escalation decision
- Evidence calculations and checked guidance
Assigned public passages and original model guidance were checked. No learner work, paid course or full licensed standard was assessed; the remaining-needs section controls any closure claim.
1. Read with a purpose.
NIST · NIST IR 8286r1: enterprise risk direction, appetite and tolerance
FREE PUBLIC READINGRead: NIST IR 8286r1, viewer p.16 (printed p.7), Enterprise Risk Direction: appetite and tolerance paragraphs, footnote 11 and the email-availability example only.
Learn: Distinguish broad leadership risk direction from the narrower operational tolerance used for a specific objective; wording varies across NIST publications.
Exact scope and exclusionsPublic official PDF readable without account or payment; selected passage independently checked on 29 October 2026.
NIST · NIST AI RMF Playbook: purpose, alternatives and go/no-go boundaries
FREE PUBLIC READINGRead: MAP 1.1 About/Suggested Actions and MAP 1.5 About/Suggested Actions only.
Learn: Start with the purpose, people affected and a non-AI alternative. Then document what risk can be accepted and what causes a no-go decision.
Exact scope and exclusionsSelected HTML text was publicly readable without payment or sign-in.
NIST · NIST AI RMF Playbook: risk tolerance and accountable roles
FREE PUBLIC READINGRead: GOVERN 1.3; GOVERN 2.1 About/Suggested Actions; GOVERN 2.3 outcome.
Learn: Connect organizational tolerance to resources and decision authority. A developer, data owner, independent reviewer and risk acceptor have different jobs.
Exact scope and exclusionsSelected HTML text was publicly readable without payment or sign-in.
UK National Cyber Security Centre and international partners · NCSC secure AI design: assess the need, data and control trade-offs
FREE PUBLIC READINGRead: Selected threat-modelling and security/functionality design headings.
Learn: Compare data exposure, action permissions, usefulness and oversight when choosing controls.
Exact scope and exclusionsSelected full heading text publicly readable without payment or sign-in.
CISA National Cyber Exercise Program / JCDC · CISA JCDC AI Cyber Tabletop Exercise: scenario and facilitator outcomes
FREE PUBLIC READINGRead: p.1 purpose; p.3 Day 2 questions and Desired Discussion Outcome. Attempt the discussion before looking at the editorial worked case.
Learn: Use a publisher discussion prompt to question risk priorities and lifecycle evaluation; its desired outcome is a facilitation aid, not an answer key.
Exact scope and exclusionsExact public PDF fetched by ordinary HTTP with browser User-Agent: HTTP 200, 342286 bytes, 10 pages. All pages read in memory. Web reader separately returned 403; access may vary.
2. Try the fictional case.
A small booking studio decides whether to pilot a drafting assistant
Fictional paper exercise. Cedar Studio answers 40 routine booking enquiries each week. Staff currently spend 8 minutes on each reply. The proposal is an approved SaaS assistant that drafts public-FAQ replies; staff continue to make bookings and approve every outgoing message. The team expects a reviewed draft to take 5 minutes. A separate booking system holds personal records. The untested proposal to upload its whole customer export is included as an option to reject or escalate. All times, counts, thresholds, policies and roles below are invented learning choices, not publisher mandates or measured business outcomes. Use no real customer data.
Your tasks
- Write the intended use and prohibited uses. Name the AI service, inputs, outputs and people who could be affected. Include a manual alternative.
- Set a fictional appetite statement and measurable pilot conditions before seeing results. Decide who may approve deployment and who may stop the pilot.
- Compare full-record uploads, public-FAQ-only drafting and the manual workflow. Select protective controls, name their owners and explain one trade-off for each.
- Review the supplied fictional test results: the first prompt produced 6 materially wrong FAQ drafts out of 40; a revised FAQ-grounded prompt produced 1 out of 40. No private-record fixture was exposed in either test. Decide what this supports and what it does not.
- Complete a decision record with benefit, residual harms, evidence limitations, approval/escalation and a review date. Compare it with the guidance only after making your decision.
3. Compare your reasoning.
Decide and write your explanation first. Where a table is wider than your screen, focus it and use the arrow keys, or scroll it sideways.
Open explained model guidance
Step 1 - Bound the use. Asset/use: FAQ Draft Assistant, supplier tenant CEDAR-TRAINING, public FAQ v12 and system prompt v7. Input: approved public opening times, prices and synthetic examples. Output: an internal reply draft with a source pointer. It may not access customer exports, set individual prices, rank customers or send mail. Customers can be harmed by wrong terms, disclosure or unfair/inaccessible responses. The manual FAQ/template workflow remains available. This is an original application of MAP 1.1, not an official NIST case.
Step 2 - Distinguish appetite, tolerance and a release test. Risk appetite is the broad amount and type of risk an organization is willing to take to pursue its objectives. Risk tolerance concerns the remaining risk it is prepared to bear after response, interpreted in the relevant use context; narrower operational limits can express that direction. Cedar's fictional appetite is to use AI only where routine drafting benefits can be obtained with low information-quality risk and strong confidentiality and human oversight. It authorizes no confidential-record uploads or unreviewed sends. Cedar operationalizes this direction through a fictional pilot tolerance/release gate: at most 1 materially wrong draft per 40 predefined standard FAQ cases, every message human approved, no prohibited-data fixture disclosed and no autonomous action. These gates are chosen controls on this pilot, not universal tolerances or estimates of population risk. The content lead owns the public FAQ data; the application owner operates the service. A security reviewer separate from the application owner checks test/control evidence with the privacy lead. The director accepts residual business risk after those reviews; the operations manager can suspend the pilot immediately. Some NIST documents use tolerance more broadly; the appetite/tolerance distinction here follows NIST IR8286r1, printed p.7.
Step 3 - Use an owner/control record.
| Risk or decision | Selected control and owner | Trade-off or residual issue |
|---|---|---|
| Confidential information sent to supplier | Privacy lead approves an explicit input allowlist; application owner disables customer-system connectors | Less personalization; users may still paste private text, so train staff and check handling |
| Wrong or invented booking terms | Content lead maintains the approved FAQ; named staff check every draft against it | Human review takes time and can miss errors |
| Unauthorized action | IT owner removes send/write permissions and uses a separate restricted identity | Additional integration work; capability restrictions do not establish factual accuracy |
| Unreviewed model/prompt change | Application owner records supplier/model identifier, prompt version and change-test evidence | Provider notice/version controls must be verified; use manual fallback when an unassessed change cannot be avoided |
| Unfair or inaccessible reply | Operations lead tests equivalent enquiries and plain-language/accessibility needs, and offers human help | Small samples cannot establish fairness for all customers |
| Sensitive logs | Security lead keeps necessary version/event metadata with restricted access and a chosen retention period | Less raw content can make later investigation harder; collection must be proportionate |
Step 4 - Interpret the evidence. Initial material-error frequency is 6/40 = 15%; revised frequency is 1/40 = 2.5%, a reduction of 12.5 percentage points on these supplied cases. The revised candidate meets Cedar's chosen FAQ gate. Zero observed leaks in these fixtures does not show leakage is impossible. A small reused test set can be unrepresentative, and human review remains a fallible control. The expected saving is (8 - 5) x 40 = 120 staff-minutes, or 2 hours per week before licence/admin costs; it is a planning estimate, not validated benefit.
Step 5 - Record the decision. Decision ID AI-ADOPT-01: conditionally approve a two-week FAQ-only pilot with all controls evidenced; keep full customer-export upload outside authorization and escalate any requested data expansion to the privacy lead and director. Owner: operations manager. Approver/risk acceptor: director, after security/privacy review. Residual risks: remaining FAQ errors, staff over-trust, unequal performance and supplier changes. Review on day 14 or sooner after a prohibited disclosure, unauthorized action or failed pilot gate. If controls cannot be demonstrated, retain the manual workflow. CISA p.3 adds a useful discussion about evaluation throughout the lifecycle; it does not approve this decision.
Original editorial source-informed application with fictional model guidance. It is not an official ISC2/PECB quiz, an independently marked learner submission, legal advice, a production approval or validated competence. Alignment is introductory CC AI-domain application only.
4. Check the work, not just the score.
- Could another reader identify the purpose, prohibited uses, exact data flow and manual alternative from your record?
- Did you distinguish the person operating the service, the data owner, the tester/reviewer and the person authorized to accept residual risk?
- Did each selected control address an identified harm and include an owner, trade-off and way to check it?
- Did you calculate 15%, 2.5%, 12.5 percentage points and 120 minutes correctly, without presenting the small test as a guarantee?
- Did your decision state conditions, remaining risk, escalation, stop authority and a review trigger?
These checks help you revise an explanation. They do not certify mastery or predict an examination result. Ask a qualified reviewer to assess a fresh case when independent feedback is needed.
5. Keep the limits in view.
Free coverage added
- A complete introductory adoption decision connects the use, asset, owners, affected people, risk appetite, data/model protection, privacy/ethical harms, control trade-offs and residual-risk decision.
- Original worked decision record, model guidance and self-check are available without an account or paid tool.
- CISA supplies an additional publisher discussion and desired discussion outcome; it does not individually assess the learner.
Still missing or outside this guide
- Independent review/marking of the learner's own fresh decision record is not provided by these public readings or the editorial model answer.
- A real adoption still needs organization-specific facts, applicable requirements, supplier evidence, affected-party input and authorized approval. Complete AI RMF/ISO 42001/legal instruction remains outside the introductory target.
Payment and access
No payment is needed for these selected readings or the paper case. Paid current teaching/marked support is optional, with exact coverage and actual feedback to be verified before purchase; no paid AI credential is required by this selection.
Open this area in the gap register