Key takeaways
- Separate effective requirements from proposed standards, staff plans and control guidance.
- Check which entity and reporting period each requirement covers before putting it in the close plan.
- Retain the framework and source versions used, and revisit the decision when a listed development changes.
No auditing standard yet sets an AI-specific procedure for the close: neither the current nor the amended text of AS 2201, the PCAOB’s standard for auditing internal control over financial reporting, refers to artificial intelligence, and PCAOB staff guidance on it is still under consideration. What is moving is the framework around it. The next binding date is December 15, 2026, when new quality-control rules for audit firms and amendments to AS 2201 take effect. Proposed international audit standards would reach system-generated reports and reconciliations, and audit committee chairs are already asking their auditors about AI.
For the controller, the useful distinction is between a requirement that changes the next reporting cycle and guidance that helps test an existing control.
Each record below gives the date that governs, who published it, what happened, why it matters, related reading and our recommended next step. Records are grouped by importance to reporting and assurance: the UK declaration and the December 15 PCAOB changes first, then proposals and guidance, with the terminology change last. Every record in this edition comes from an official source: a standard setter, regulator, framework body or government.
UK Code: evidence for material-control declarations
- Date that governs: January 1, 2026; Provision 29 of the UK Corporate Governance Code 2024 applies to financial years beginning on or after that date, while the rest of the 2024 Code has applied to financial years beginning on or after January 1, 2025.
- Source tier: Official (Financial Reporting Council, United Kingdom).
- What happened: The Code applies to companies listed in the commercial companies or closed-ended investment funds categories, which must apply its Principles and comply with, or explain against, its Provisions. Provision 29 asks boards to make a declaration on the effectiveness of their material internal controls. The FRC’s guidance says the board decides which controls are material, and that these could include controls over information and technology risks, including new technologies such as artificial intelligence. The guidance does not set out a framework or define a material control, and neither the Code nor the guidance requires or expects external assurance on the declaration. On September 23, 2026 the FRC published an information sheet on cyber security, which says many companies are likely to treat it as a material control, given their reliance on digital systems and technologies such as AI.
- So what: For a company reporting against the 2024 Code with a calendar financial year, the first year in scope ends December 31, 2026. Where a board treats an AI-dependent close control as material, management will need evidence that the control worked, not only that it exists. That evidence is the same kind the IAASB proposals below point to: what the AI received, what it produced, and what review could have caught.
- Related reading: Review Is a Control Only When It Can Catch the Error; Internal Control Over Generative AI, Applied to the Close.
- Our recommendation: Calendar-year companies reporting against the Code should decide whether any AI-dependent close or reporting control is material, and confirm the evidence supporting the declaration. Recommended by December 31, 2026, the first year-end in scope.
- Sources: FRC, “UK Corporate Governance Code 2024”, January 22, 2024; FRC, “Corporate Governance Code Guidance”, January 29, 2024; FRC, “Provision 29 Mythbuster: Focus on Cyber”, September 23, 2026.
PCAOB: audit-firm quality controls change on December 15
- Date that governs: December 15, 2026, the effective date of QC 1000, of the 2026 amendments to it if the SEC approves them, and of the related amendments to AS 2201.
- Source tier: Official (PCAOB and SEC, United States).
- What happened: On September 9, 2026, the PCAOB adopted amendments to its quality-control standard, QC 1000, which do not change its December 15, 2026 effective date and are subject to SEC approval. The amendments include rescinding the requirement for an External QC Function, allowing firms to choose the date of their annual QC evaluation, and shortening the retention period for QC documentation from seven years to five. The SEC issued notice of the filing on September 15, 2026, published in the Federal Register on September 18, with comments due October 9, 2026; approval remains pending. Commenters asked the PCAOB for guidance on quality-control matters including due diligence on AI tools used on engagements. Separately, amendments to paragraph .09 and a new paragraph .99 of AS 2201 have been adopted and approved by the SEC and take effect on December 15, 2026; paragraphs .26 and .27 (the period-end financial reporting process) and .42 through .47 (testing controls) are not among them. New paragraph .99 says an auditor who determines, after issuing the ICFR report, that an engagement deficiency exists should act under AS 2901, unless it is probable the report is not being relied upon.
- So what: These changes govern how audit firms run and evidence their own quality systems, not what management must do over its controls. For a finance team using AI in the close, the auditor’s ICFR requirements on the period-end process and on testing controls carry over unchanged, so an AI-dependent control will be judged under the same standard as any other. From December 15, 2026, the auditor’s broader responsibility to respond to deficiencies found after the report is issued can bring follow-up questions to management.
- Related reading: Review Is a Control Only When It Can Catch the Error; Treat AI Output Like an Unsupported Journal Entry.
- Our recommendation: Ask the engagement partner whether the firm’s QC 1000 implementation changes how it plans or supervises your ICFR audit, including which AI tools the engagement team will use on your data. Recommended by December 15, 2026. This page will recheck the SEC approval and the AS 2201 text on December 16, 2026.
- Sources: PCAOB, “PCAOB Adopts Amendments to Its Quality Control Standard”, September 9, 2026; SEC, Release No. 34-106372, notice of filing of proposed amendments to QC 1000, File No. PCAOB-2026-01, September 15, 2026; PCAOB, AS 2201, An Audit of Internal Control Over Financial Reporting, as amended, effective December 15, 2026; PCAOB, Release No. 2024-005, A Firm’s System of Quality Control and Other Amendments, May 13, 2024.
PCAOB: AI guidance remains under consideration
- Date that governs: September 30, 2026, when the PCAOB updated its standard-setting, research and rulemaking agendas after seeking public comment on them for the first time.
- Source tier: Official (PCAOB, United States).
- What happened: PCAOB staff lists guidance for auditors on the impact on the audit of a company’s use of AI in financial reporting, and on the auditor’s own use of it, among the topics it is still considering, not guidance it has issued. In October 2026 the entry was retitled “Use of Super Intelligence”, and the agenda page now describes super intelligence among the technologies auditors and public companies increasingly use; the PCAOB has not said why. The PCAOB also lists a research project on data and technology. Its staff publications page lists no publication on AI since its July 2024 generative AI spotlight. In April 2026, PCAOB staff reported that, in its 2025 conversations with more than 250 audit committee chairs, chairs were increasingly seeking their auditors’ views on AI’s potential effects on financial reporting and on the controls in place.
- So what: Until guidance is issued, auditors of US public companies will assess AI in the close under existing standards, and expectations may differ between firms. The questions audit committee chairs report asking give controllers a reason to have their own account of where AI touches the close before the auditor asks.
- Related reading: Internal Control Over Generative AI, Applied to the Close; Treat AI Output Like an Unsupported Journal Entry.
- Our recommendation: Agree with the audit committee and the external auditor on an inventory of close activities that use AI and the evidence kept for each. Recommended by December 31, 2026. This page will recheck the PCAOB agenda by January 19, 2027.
- Sources: PCAOB, “PCAOB Updates Standard-Setting, Research, and Rulemaking Agendas”, September 30, 2026; PCAOB, “Standard-Setting, Research, Rulemaking, and Related Activities”, updated September 30, 2026; PCAOB staff, “Spotlight: 2025 Conversations With Audit Committee Chairs”, April 2026.
IAASB: proposed audit-evidence changes would reach system outputs
- Date that governs: December 15, 2026, the comment deadline for the IAASB’s proposals, published on August 5, 2026.
- Source tier: Official (IAASB, international).
- What happened: The IAASB opened a public consultation on proposed revisions to ISA 330, ISA 500 and ISA 520, which it describes as modernizing the standards for a technology-enabled audit while remaining principles-based rather than technology-specific. Proposed ISA 500 (Revised) would require the auditor, when information produced by the entity is used as audit evidence, to evaluate its accuracy and completeness to the extent necessary for the purpose of the procedure. The IAASB describes that requirement as equivalent to the current standard; the new application material counts outputs of the entity’s information system as information produced by the entity and lists reconciliations and system-generated reports among the examples. The exposure draft also names automation bias, placing undue weight on output from automated systems, as an example of bias. The IAASB anticipates approving the final standards in December 2027, effective for audits of periods beginning at least 18 months later. Separately, its technology quality-management workstream, launched in June 2025, discussed drafts of its first two guides on AI-enabled tools in September 2026, and the first is to be considered at a meeting on October 28, 2026.
- So what: For audits performed under the ISAs, an AI-prepared reconciliation or a system-generated close report would, as drafted, fall within information produced by the entity, so the auditor’s duty to evaluate its accuracy and completeness would reach it. Nothing changes until the revised standards are approved and effective, but finance teams that already keep evidence of what an AI tool received and what it may have missed will be ready either way.
- Related reading: Treat AI Output Like an Unsupported Journal Entry; Review Is a Control Only When It Can Catch the Error.
- Our recommendation: Organizations that respond to IAASB consultations should comment by December 15, 2026; others should expect no final standard before December 2027. This page will recheck the project and the October 28 meeting by January 19, 2027.
- Sources: IAASB, “Strengthening the Risk-Based Audit Framework: Proposed Revisions to ISA 330, ISA 500, and ISA 520”, August 5, 2026; IAASB, “Exposure Draft, Proposed ISA 500 (Revised), Audit Evidence”, August 2026; IAASB, “Exposure Draft for the Audit Evidence and Risk Response Project Overall”, August 2026; IAASB, “IAASB eNews October 2026”, October 5, 2026.
FRC: ask how the auditor uses AI on your data
- Date that governs: March 30, 2026, the publication date of the FRC’s guidance.
- Source tier: Official (Financial Reporting Council, United Kingdom).
- What happened: On March 30, 2026, the FRC published guidance for audit firms on using generative and agentic AI tools in audit engagements, its second guidance on AI in audit, and said it codifies good practice rather than responding to identified quality deficiencies. The FRC states that regulatory accountability for deploying AI tools and for the quality of audit outputs is unchanged, and that the human auditor remains accountable. The guidance defines agentic AI systems as systems that can orchestrate and execute multiple components or tasks toward a goal with some degree of autonomy. It notes a cybersecurity risk if audited entity data is input into cloud-based AI applications, but limits its own attention to risks to audit quality. Among design mitigations, it describes scope and authority boundary rules that define what actions a system can perform and when human authorization is required.
- So what: Your auditor’s AI tools may process your close data, and this guidance addresses audit quality rather than how the auditor protects that data. Its boundary rules mirror the questions finance should ask of its own agents: what the system may do, and where a human must authorize the next step.
- Related reading: Your AI Agent Is an Unlisted Signer; Review Is a Control Only When It Can Catch the Error.
- Our recommendation: Ask your auditor which AI tools will process your data in the next audit, where that data is processed, and what authorization governs it. Recommended by December 31, 2026.
- Sources: FRC, “Innovative new guidance supports audit firm adoption of emerging AI technologies”, March 30, 2026; FRC, “Generative and Agentic AI Guidance”, March 2026.
COSO: retain the guidance file you rely on
- Date that governs: February 23, 2026, the release date of the COSO-commissioned publication.
- Source tier: Official (COSO, United States; private-sector framework body).
- What happened: COSO released Achieving Effective Internal Control Over Generative AI on February 23, 2026. The publication states that generative AI outputs can be confidently wrong and that controls should treat them as claims requiring validation. It says that when management relies on AI outputs as evidence for an ICFR control, the AI control patterns should meet the same evidence standards expected for ICFR. COSO’s page links two files, created February 19, 2026 and September 1, 2026, which differ in one Appendix B heading and one typo, with no version notice. COSO’s news page lists no later publication on generative AI. The publication describes itself as general information whose views may differ from those of regulators.
- So what: COSO’s publication adapts the 2013 internal control framework to generative AI. It is the control-oriented publication our close research relies on, and because two versions are live, control documentation should record which file it cites.
- Related reading: Internal Control Over Generative AI, Applied to the Close; Treat AI Output Like an Unsupported Journal Entry; Your AI Agent Is an Unlisted Signer.
- Our recommendation: Retain the COSO file your control documentation relies on, and record which of the two versions it is. Recommended by December 31, 2026. This page will recheck it by January 19, 2027.
- Sources: COSO, “Achieving Effective Internal Control Over Generative AI”, February 23, 2026.
NIST: keep control mappings tied to a version
- Date that governs: April 8, 2026, the latest update date on NIST’s record for its generative AI profile, NIST AI 600-1.
- Source tier: Official (NIST, United States; voluntary framework).
- What happened: NIST lists AI 600-1 as published July 26, 2024 and updated April 8, 2026, with no notice that it has been superseded or withdrawn, and does not say what the update changed. NIST states that the AI Risk Management Framework 1.0, released January 26, 2023, which the profile accompanies, is being revised as part of the White House AI Action Plan.
- So what: Finance teams that map AI controls to NIST’s functions or to AI 600-1’s risk categories, such as confabulation and automation bias, are mapping to a framework under revision. The mapping still holds, but it should be dated and rechecked when NIST publishes the revised framework.
- Related reading: Review Is a Control Only When It Can Catch the Error; Treat AI Output Like an Unsupported Journal Entry.
- Our recommendation: Record the NIST versions your AI control documentation cites (AI RMF 1.0, January 26, 2023; AI 600-1, July 26, 2024, updated April 8, 2026). Recommended by December 31, 2026. This page will recheck NIST’s status by January 19, 2027.
- Sources: NIST, “AI Risk Management Framework”, January 26, 2023; NIST, “Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile,” NIST AI 600-1, July 26, 2024, updated April 8, 2026.
EU AI Act: literacy now, listed high-risk uses later
- Date that governs: August 2, 2026, from which the EU AI Act applies generally; its AI literacy provisions first applied on February 2, 2025 and were amended in July 2026.
- Source tier: Official (European Union).
- What happened: The Act’s AI literacy duty, as amended by the Digital Omnibus published in July 2026, requires providers and deployers to take measures to support their staff’s AI literacy, without guaranteeing any specific level. Annex III lists as high-risk AI systems intended to evaluate the creditworthiness of individuals or establish their credit score, except systems used to detect financial fraud. Since the Omnibus, the high-risk obligations apply to Annex III systems from December 2, 2027. None of the Annex III uses is a close or consolidation activity.
- So what: For most finance teams in the EU, the duty that applies now is AI literacy for the people who use AI in their work. Finance functions in lenders, and any that score individuals’ creditworthiness, have until December 2027 to meet the high-risk requirements.
- Related reading: What Good AI Use Looks Like at Each Level of a Team; What We Mean by AI.
- Our recommendation: EU-based finance teams should record what they do to support staff AI literacy; lenders should confirm whether any credit-scoring model falls within Annex III. Recommended by December 31, 2026.
- Sources: European Union, Regulation (EU) 2024/1689 (Artificial Intelligence Act), June 13, 2024; European Union, Regulation (EU) 2026/1744 (Digital Omnibus on AI), July 8, 2026.
Terminology watch: federal use of “super intelligence”
- Date that governs: September 29, 2026, when Executive Order 14434 was signed; it was published in the Federal Register on October 2, 2026.
- Source tier: Official (Executive Office of the President, United States).
- What happened: The order makes it the policy of the Administration that, to the maximum extent permitted by law, the executive branch uses the terms “Super Intelligence” and “SI” in place of “Artificial Intelligence” and “AI”. It directs agencies to do so in correspondence, public communications, websites, reports and other non-statutory documents. It defines the new terms as the technologies covered by the existing statutory definition of artificial intelligence, and asks for proposed legislative language for a federal definition within 60 days. NIST’s record for its generative AI profile now uses “SI” in its topic tags, while the publication itself is unchanged.
- So what: The change is one of terms, not of substance: the definition it points to is the existing one. Finance teams that cite federal frameworks in control documentation should expect the new wording in updated pages and record which version they cite. The order is addressed to executive departments and agencies and does not name the PCAOB or the SEC.
- Related reading: What We Mean by AI.
- Our recommendation: No action is needed beyond noting the change in terminology; this page will recheck for the proposed statutory definition by January 19, 2027.
- Sources: Executive Office of the President, Executive Order 14434, “Inaugurating the Era of Super Intelligence,” 91 FR 63129, September 29, 2026; NIST, “Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile,” NIST AI 600-1, July 26, 2024, updated April 8, 2026.
How to use these records
The records are not legal or audit advice. They are ordered by how directly each affects reporting and assurance, not by legal force. Dates labeled “Recommended by” are our suggested timing, not legal deadlines. Proposals, staff plans and framework publications are labeled as such; only an adopted and effective standard, or an applicable code provision (which listed companies must apply or explain departures from), sets requirements. Discuss how any of these developments applies to your organization with your auditors and advisers.
Review status
Active Watch: reviewed quarterly and whenever a listed development changes, such as a final standard, an approval or a new publication. Readers who know of a development we have missed can tell us through Help Improve.


