Governance ARA-261031-2D523F

Library / ARA-261031-2D523F / governance

Worked example: Generative AI chatbot for internal HR policy

Versions of this assessment, reviewer sign-off, suggested framework mapping, and every linked deliverable.

Sign-off

Sign off version 1 (the latest run). 15 of 15 items are not yet reviewed; the sign-off will record this.

Versions (1)

Version	Saved (UTC)	Model	SHA-256
1	2026-10-31 06:37:15	Worked example: prepared in advance, not generated live	f73afa28ce4bdd86

Client

Item-by-item review

Confirm each AI-drafted risk, threat, regulatory finding and control. 0 of 15 reviewed (0 changed, 0 rejected). "Accepted with changes" needs a note describing the change. The counts are sealed into the sign-off.

Type	ID	Item
risk	R-01	Employees see colleagues' salary or medical data (Critical)
risk	R-02	Wrong policy answers (Medium)
risk	R-03	External LLM retains HR data (High)
risk	R-04	Translation errors change policy meaning (Medium)
threat	T-01	Permission bypass through retrieval (Critical)
threat	T-02	Indirect prompt injection in a policy document (Medium)
regulation	REG-01	Privacy Act 1988: potentially_applicable
regulation	REG-02	General Data Protection Regulation (EU) 2016/679: potentially_applicable
regulation	REG-03	Regulation (EU) 2024/1689 (AI Act), Article 50 transparency: potentially_applicable
regulation	REG-04	Personal Information Protection Law (China); LGPD (Brazil): not_enough_information
control	C-01	Permission-aware retrieval
control	C-02	Exclude personal records from the index
control	C-03	Provider terms: no training, limited retention
control	C-04	Grounded answers with citations and version control
control	C-05	Leakage testing before go-live

Deliverables

Board pack (Word)Risk register (Excel) Printable report

Failure simulations	none
Operating envelopes	none
ADM inventory entries	none (add)
Vendor reviews	none (start)

Framework mapping (suggested)

Suggested by keyword rules at heading level; verify each mapping against the licensed standard before using it in a statement of applicability or audit.

Control	ISO/IEC 42001 Annex A	NIST AI RMF	ISO/IEC 27001:2022	OWASP LLM 2025
C-01 Permission-aware retrieval	-	-	-	LLM06 Excessive Agency; LLM08 Vector and Embedding Weaknesses
C-02 Exclude personal records from the index	A.2 Policies related to AI	-	-	-
C-03 Provider terms: no training, limited retention	A.10 Third-party and customer relationships	-	-	-
C-04 Grounded answers with citations and version control	A.2 Policies related to AI; A.6 AI system life cycle	GOVERN	5 Organizational	LLM09 Misinformation
C-05 Leakage testing before go-live	A.6 AI system life cycle	MEASURE	-	LLM08 Vector and Embedding Weaknesses

Headings with no mapped control

ISO/IEC 42001 Annex A: A.3 Internal organization; A.4 Resources for AI systems; A.5 Assessing impacts of AI systems; A.7 Data for AI systems; A.8 Information for interested parties of AI systems; A.9 Use of AI systems

NIST AI RMF: MAP; MANAGE

An uncovered heading is a prompt to check whether a control is needed, not proof of a gap.

Decision support only, not legal or regulatory advice. AI-generated ratings and citations must be reviewed by a qualified person before they are relied on.