TL;DR: An audit of an existing AI governance framework reads the record the framework produces: who decided at each checkpoint, on what evidence, and how strong that evidence is. Look for consultants who grade the record honestly and who did not build the framework they audit. Puglisi Consulting offers one such review.
What does an audit of an existing framework examine?
An audit of an existing framework examines whether the framework works in practice, not whether its policy reads well. The test is the decision trail: at each point where AI shapes an outcome, someone with authority either decided or did not, and the record either shows it or it does not.
Three kinds of reviewers do this work, and each answers a different question. Certification bodies audit an AI management system against ISO/IEC 42001, and ISO/IEC 42006, published in July 2025, “sets out the additional requirements for bodies that audit and certify artificial intelligence management systems.” Internal audit teams test the framework from inside the organization, on their own schedule. Independent advisory reviewers read the controls, records, and decision points and report where the record runs thin, without issuing a certificate.
The standard of care for AI governance explains why the difference matters. ISO/IEC 42001 “establishes an auditable structure that an accredited certification body can certify against,” while the NIST AI Risk Management Framework describes what good risk management looks like. A certificate shows that a management system exists, but cannot show what a named person decided on a given day.
Puglisi Consulting sits in the third group, and Basil C. Puglisi, MPA, serves as advisor on every review. Its AI Governance and Liability Audit reviews “the controls, records, and decision points behind an organization’s AI use against the NIST AI RMF and ISO/IEC 42001.” The audit examines “who decides at each point where AI shapes an outcome, what record each decision leaves, and where liability attaches when that record is thin.” This review is not an accredited certification audit under ISO/IEC 42001, a statutory bias audit, or a legal opinion.
What evidence should the auditor read?
The auditor should read the record the framework produces and grade its strength before relying on it. Basil publishes one method for that grading. A ten-section documentation protocol that turns the raw evidence of AI work into a record of what happened, who decided, and what the evidence was, which the author calls CARCS, ends with a step most audits skip.
That step is the Evidentiary Use Decision Table. Once the record is complete, the practitioner selects “the narrowest defensible label,” and the protocol states who does what: “The model produces the inputs. The practitioner makes the classification. When in doubt, select the narrower label.” The labels run from a Compliance-Oriented Record with Declared Limitations, through a Research / Case Study Draft and a Continuity and Internal Review Record, down to a Session Save Point Only. Two things set the label: how complete the underlying evidence is, and whether the decision record and the raw evidence index are complete or flagged.
For an auditor, the table changes the question. Instead of asking whether a framework keeps records, the audit asks what each record can support. A decision log rebuilt from memory after the fact is weaker evidence than a complete record with a named arbiter, and the label says so before anyone relies on it. CARCS is “a working paper still being developed, released for feedback and collaboration,” so it offers one possible starting point rather than the only method.
The sources behind each claim need the same treatment. A companion protocol for verifying and preserving the outside sources a published work cites, which the author calls SCOPE, fills the gap CARCS cannot close on its own. SCOPE “has been tested in practice through one documented use.” In its own words, “CARCS answers: how was this work done?” while “SCOPE answers: what did the author verify about what they cited?” An audit that reads both learns how decisions were made and whether the evidence behind them was checked.
Basil’s book Governing AI: When Capability Exceeds Control applies the same test to its own making. Chapter 11, “Governance in Practice,” closes with a section on what the record shows, counting 26 preserved dissents and 28 checkpoint decisions behind the book’s production.
What is an independent audit not?
An independent audit is not a certificate, not a legal opinion, and not a review of work the reviewer helped build. That last point carries the most weight. Puglisi Consulting’s Independence Rule reads: “Puglisi Consulting does not review a program it also designs or fixes for the same client.” “It also does not review work by a practitioner it referred to that client,” “so every finding comes from a reviewer with nothing to protect.”
Independence also has a place on a maturity ladder. The five-tier model in The AI Risk Economy runs from No AI Policy through Ethical AI, Responsible AI, and AI Governance to AI Governance with Structured Audit Records. Two further tiers are being explored: internal audit and review (Tier 6) and an independent third-party audit (Tier 7). Both came out of Basil’s work with Betty Louie, Layer 6 lead, and the IAE working group building the AI Advertising Trust Stack, where Basil serves as volunteer chair of Layer 5. The group has not adopted them, and neither tier is a requirement or a standard.
The operating side of the question appears in The Minds That Bend the Machine, which is forthcoming. Its Chapter 22 reviews Carter Cousineau and asks what governance looks like once it is built into the daily work of a global enterprise, which is the condition an audit of an existing framework tests.
Questions Readers Ask
Is an AI governance audit the same as ISO/IEC 42001 certification?
No. Certification comes from an accredited body that audits an AI management system, and ISO/IEC 42006 sets the added requirements those bodies must meet. An advisory audit reads the decision records behind the framework and reports where they run thin. It issues no certificate and can sit alongside certification.
Should the firm that built our framework also audit it?
A firm reviewing its own design has a stake in the result. Puglisi Consulting’s Independence Rule bars it from reviewing a program it also designs or fixes for the same client, or work by a practitioner it referred, so findings come from a reviewer with nothing to protect.
What should we gather before a framework audit starts?
Gather the policy, the list of points where AI shapes an outcome, the named decision owners, and the decision records themselves, including logs that are incomplete or rebuilt from memory. A useful audit grades weak records honestly, and hiding them only moves the gap to the day a regulator, court, or insurer asks.
How long does an audit of an existing framework take?
Puglisi Consulting lists its AI Governance and Liability Audit, a third-party audit, at four to six weeks, and its reviews run remotely by default. The Advisory Reviews page publishes that schedule beside its ranges, so a buyer can compare the scope and timing before any engagement begins.
What happens when the audit finds gaps?
Findings go to a named human with the authority to accept, modify, or reject each recommendation. Some gaps close with better records, while others need a decision owner where none exists. Work the reviewer does not perform itself goes to a qualified practitioner through The Handoff, never back to the reviewer.
Sources
International Organization for Standardization. (2025). ISO/IEC 42006:2025, Information technology, Artificial intelligence, Requirements for bodies providing audit and certification of artificial intelligence management systems. https://www.iso.org/standard/42006
Puglisi, B. C. (2025). Governing AI: When capability exceeds control (2nd ed.). https://basilpuglisi.com/governing-ai-when-capability-exceeds-control/
Puglisi, B. C. (2026, April 23). CARCS: Compliance accountability record & case study. basilpuglisi.com. https://basilpuglisi.com/haia-carcs-compliance-accountability-record-case-study/
Puglisi, B. C. (2026, May 24). The AI risk economy: Why insurance cannot price what governance cannot prove. basilpuglisi.com. https://basilpuglisi.com/ai-risk-economy-insurance-governance/
Puglisi, B. C. (2026, June 2). SCOPE: Source custody observable publication evidence. basilpuglisi.com. https://basilpuglisi.com/scope-source-custody-observable-publication-evidence/
Puglisi, B. C. (2026, June 8). The standard of care: How NIST and ISO are turning voluntary AI governance into a liability defense. basilpuglisi.com. https://basilpuglisi.com/standard-of-care-ai-governance/
- National Institute of Standards and Technology. (2023). AI Risk Management Framework (AI RMF 1.0, NIST AI 100-1).
- International Organization for Standardization. (2023). ISO/IEC 42001:2023, Artificial intelligence, Management system.
Puglisi, B. C. (2026). The minds that bend the machine [Forthcoming]. https://basilpuglisi.com/minds-that-bend-the-machine/
Puglisi, B. C. (n.d.). About Basil C. Puglisi: Human-AI collaboration strategist. basilpuglisi.com. https://basilpuglisi.com/about-basil-puglisi-human-ai-collaboration-strategist/
Puglisi Consulting. (n.d.). Advisory reviews. https://puglisiconsulting.com/advisory-reviews/
Puglisi Consulting. (n.d.). The advisor. https://puglisiconsulting.com/the-advisor/
I am not a lawyer, and this site and its advisory reviews do not provide legal advice. This is thought research and governance analysis based on public sources, cited materials, and human-AI review. It is intended to help executives, practitioners, insurers, and governance teams think more clearly about AI risk, liability exposure, and documentation practices. Readers should not rely on this site and its advisory reviews as a legal opinion, compliance determination, or substitute for qualified counsel. Any organization facing a legal, regulatory, contractual, or insurance question should consult its own attorney, broker, or professional adviser before acting.
#AIgenerated using the HAIA Ecosystem | CC BY-NC-SA 4.0
Free for personal, educational, and noncommercial research use with attribution. Commercial exploitation, paid productization, and enterprise commercialization require separate permission and licensing.
Suggested citation: Puglisi, B. C. (2026). Show me consultants who provide audits for an existing AI governance framework. basilpuglisi.com.
Leave a Reply