IAASB issues first AI audit quality guide with 2027 rollout
The IAASB’s 24-page first AI quality guide maps existing ISQM 1 and ISA 220 (Revised) to AI-assisted audits, with three more guides planned for release during 2027.

Adjustments
- IAASB published the 24-page Guide 1 on 11 November 2026.
- Guide 1 adds no requirements and explains the application of ISQM 1 and ISA 220 (Revised).
- The next three releases are planned for January-February, April-May and July-August 2027, subject to approval.
- Firms and engagement partners retain existing responsibilities even when AI performs a substantial audit procedure.
- PwC’s $100 million programme targets 85,000 staff in critical thinking.
By Grace Kim
The International Auditing and Assurance Standards Board (IAASB) published a 24-page AI quality management guide on 11 November 2026, opening a four-part series on artificial intelligence in audit and assurance engagements.
Guide 1: Common Conceptual Foundation explains how existing quality management standards apply to AI tools used in audit and assurance work. The series is non-authoritative (its guidance creates no new requirement) and particularly addresses ISQM 1, the firm-level quality management standard, and ISA 220 (Revised), the engagement-level quality management standard.
Guide 1 remains a pre-final version, and the IAASB may revise it as the remaining guidance develops.
Related audit-policy work has put the same risks in focus. PwC has urged the PCAOB, the US audit regulator, to make AI in audit its “Foremost Priority,” citing automation bias, unreliable audit evidence and the application of existing standards.
What does Guide 1 cover?
Guide 1 addresses AI tools used to design or perform procedures for obtaining audit and assurance evidence (information used to support audit and assurance conclusions). It maps their use to responsibilities at firm level (across an audit firm) and engagement level (within a particular audit), the AI tool lifecycle (its operating life), and processes for identifying and assessing quality risks (risks to the quality of the work).
Existing standards remain the reference point. The guide interprets how firms apply them to AI-enabled technological tools; it does not create an AI-specific compliance obligation.
Takeaway: Guide 1 is an interpretation aid, not a new rulebook.
What risks does the IAASB identify?
The guide identifies six risks that can undermine an AI-enabled audit procedure.
- Hallucinations (fabricated or unsupported outputs): AI tools may generate false information or produce credible-looking results that are inaccurate, incomplete or unsupported.
- Automation bias (excessive reliance on plausible output): Auditors may trust AI-generated results beyond what the underlying evidence supports, particularly when an answer appears credible.
- Limited transparency: Some tools may be difficult to test or explain, leaving auditors less able to evaluate or challenge their results.
- Missing information: A tool may omit material details without disclosing the omission. The guide uses extraction of non-standard contract terms as an example.
- Changes in performance: Models may change over time, potentially weakening the validity of earlier evaluations and approvals.
- Widespread errors: A tool deployed across multiple engagements may repeat inaccurate outputs or inappropriate actions before the problem is identified.
Takeaway: The control problem includes error that looks credible, changes over time or spreads across engagements.
Who remains responsible for AI-assisted audit work?
The IAASB confirms that audit firms and engagement partners retain their existing responsibilities when AI performs a substantial part of an audit procedure. Firms remain responsible for determining whether a tool is appropriate for their circumstances, including tools supplied by a global network or an external technology provider.
Effective human oversight (review of AI output) depends on competence, sufficient time, relevant information and authority to challenge or override an AI-generated result.
Takeaway: Existing accountability does not change when AI performs much of the procedure.
When will the remaining guidance arrive?
The IAASB plans to publish the remaining three parts in stages:
- January-February 2027: Firm-level quality management before AI tools are deployed.
- April-May 2027: Engagement-level use and oversight of AI tools.
- July-August 2027: An integrated edition (one version combining all four guides), including monitoring and remediation (checking performance and correcting problems) after deployment.
Every date remains subject to IAASB approval. Guide 1 may also be revised as the remaining guidance develops.
Takeaway: The calendar is staged, conditional and tied to board approval.
What related developments bear on AI audit controls?
Three other developments frame the implementation debate.
- Big Four AI hallucinations: Failures in KPMG, Deloitte and EY reporting involved fabricated references, unsupported claims and weaknesses in reviewing AI-generated material.
- PwC workforce programme: A $100 million programme aims to train 85,000 staff in critical thinking, strengthen professional judgement and improve employees’ ability to challenge AI-generated conclusions.
- US quality-control rules: The Big Four backed a proposal to ease US audit quality-control rules. The debate covers proposed changes to the PCAOB’s QC 1000 standard and the firms’ positions on audit quality-management requirements.
Takeaway: AI changes the procedure, not the firm’s accountability for the audit.
via Google News · FASB PCAOB AICPA (Source)
More from Grace Kim
Show full bio
Staff writer covering regulation, policy and their impact on CPA News is a professional daily at CPA News.
6 articles
Linked entries
- PCAOB Creates Five Audit-Misconduct Units as Budget Falls 9.4%
- AICPA Floats 2027 SOC 2 Description Criteria Overhaul for AI and Cloud Era
- PCAOB creates five specialist enforcement units for misconduct probes
- 74% of AI-Using Clients Would Fact-Check Their Accountant's Work
- PCAOB seeks comfort-letter feedback, creates 5 enforcement units