What Did the Big 4 Actually Tell the PCAOB About AI?
Every major audit firm that commented asked the PCAOB for guidance on AI, not formal rules, in comment letters filed this month.
Guidance, not rules. That's the short version. It's a more cautious ask than you'd expect from firms racing to build AI into their own audit work.
The Public Company Accounting Oversight Board opened the request on June 23, 2026. It was the board's first-ever public call for input on its future standard-setting priorities. Chairman Demetrios Logothetis called it a transparency move: "Transparency is essential to earning and keeping the public's trust." The request touched several topics, from fraud detection to auditor independence. But the sharpest responses came on one line item: data, technology and artificial intelligence.
The comment period closed August 7 with 33 total submissions. Going Concern's roundup of the letters, published August 12, laid out where the industry's biggest names landed.
PwC's specific asks
PwC's August 5 letter called AI "the board's foremost priority." It asked for an AI task force, near-term guidance instead of formal rules, and closer contact between the PCAOB and firms' inspection teams. The letter named four gaps the guidance should cover: data bias, explainability, automation bias, and whether AI-generated audit work can be reconstructed after the fact.
PwC was careful to draw a line, though. Its letter states that "humans should ultimately remain responsible for judgments including risk assessments, the sufficiency of audit evidence, contradictory information and the final audit conclusion." The tool can help find the answer. It doesn't get to sign off on it.
Where EY, Deloitte, KPMG and Grant Thornton agree
EY's 17-page letter went further, warning that "developing premature or overly prescriptive standards could create unintended consequences and hinder innovation." Deloitte proposed a two-phase plan instead: staff guidance now, formal standards later, timed to match international standard-setters. KPMG's letter landed in the same place, arguing that "staff guidance, supported by continued research and monitoring, is preferable to standard setting" for now.
Even Grant Thornton, a Top 25 firm and not part of the Big 4, sent nearly the same message back in June. It asked for "clear, comprehensive, principle-based guidance," not prescriptive rules. The AICPA's letter matched too: staff guidance on how existing standards apply to AI-assisted work, not a new standard written from scratch.
| Firm | What they asked for | Specific concern named |
|---|---|---|
| PwC | AI task force, near-term guidance | Data bias, explainability, automation bias, work reconstruction |
| EY | No new standards yet | Premature rules "hinder innovation" |
| Deloitte | Two-phase: guidance now, standards later | Alignment with international standard-setters |
| KPMG | Staff guidance over new standards | Research and monitoring first |
| Grant Thornton | Principle-based guidance | Consistency without rigid rules |
Six firms, six letters, one answer. Nobody who builds audits for a living asked the PCAOB to move fast.
Why Would Firms Building AI Audit Tools Ask Their Regulator to Slow Down?
Firms say AI guidance can update faster than a formal rule, so slower, principle-based direction beats a standard that goes stale by next audit season.
On its face, that sounds sensible for a technology that changes month to month. A rule written in 2026 could be obsolete by the 2027 audit season.
But guidance that isn't a rule still has to be interpreted by somebody. The firms asking for flexible, principle-based direction are the same firms with the compliance and quality-control staff built to interpret ambiguous guidance in-house. A four-partner firm reading the same PCAOB guidance doesn't have that staff. Flexibility that's easy for PwC to absorb isn't automatically easy for a firm a fraction of its size.
The Case for Moving Faster: What Happens When AI Gets It Wrong
Hofstra accounting professor Jack Castonguay argues the PCAOB should set AI audit standards now, not wait, because existing rules predate AI-enabled auditing.
Not everyone agrees the industry should get to set its own pace. Writing in Bloomberg Tax, Castonguay put it directly: "the standards weren't created for the capabilities AI-enabled auditors now possess."
His central claim: three of the four largest accounting firms have already retracted or refunded a published report after AI hallucinations were found in it. The record backs him up.
Three named incidents
KPMG withdrew its "Total Experience" report in October 2025. UBS, the UK's National Health Service, Swiss Federal Railways and Transport for London all disputed claims it made about their AI use. Of 45 references in the report, only 5 checked out.
Deloitte Australia refunded part of an AU$440,000 government contract after a welfare-compliance report it produced contained fabricated academic citations and a fake court quote. Deloitte admitted using an OpenAI model without disclosing it upfront. EY Canada withdrew a cybersecurity report after finding 16 of its 27 references were fictitious.
Those incidents predate this month's letters, and none involved a financial-statement audit. But they're the real version of the exact risk PwC named. Work that can't be reconstructed. No clean record of who checked it before it went out the door.
Reading the two positions side by side
Both sides have a real point, and neither is wrong on its own terms. The firms asking for guidance aren't wrong that a fast-moving technology resists rigid rules. Castonguay isn't wrong that "wait and see" already has a body count of retracted reports behind it, even if none of them were audit opinions. The PCAOB was built in 2002 specifically to set uniform standards ahead of the next failure rather than after it. That history cuts against "give it more time" as a complete answer, even from firms with good reasons for asking.
Free CPA AI Policy Checklist
Before staff paste client data into AI, check the rules your firm is missing.
Get the free checklist and join the Dispatch for practical AI controls, vendor questions, and client-data safeguards for accounting teams.
Free. No spam. You will also get the Nexairi Dispatch.
What Does "Guidance, Not Rules" Actually Mean for a Firm That Isn't in the Room?
Six comment letters are shaping the eventual PCAOB standard right now, and a smaller firm outside that group won't get another chance to weigh in.
PCAOB guidance likely lands sometime in 2027. When it does, expect it to lean on the same concerns these letters already put on record: data bias, explainability, traceability, human sign-off. That guidance becomes the working answer to what inspectors ask for, whether your firm helped write it or not.
That's not a reason to panic. It's a reason to build the paper trail before you're told to, using the same concerns PwC already named as your checklist. Nexairi laid out a similar five-item evidence checklist for the PCAOB's already-adopted audit standards, and the same habit applies here.
What Should Your Firm Check Before PCAOB Guidance Lands?
Three questions from the letters themselves can tell you today whether your firm's AI-assisted audit work would hold up to an inspector's request.
Can you reconstruct an AI-assisted conclusion? PwC named this specifically. Say an inspector asked you to show how an AI tool reached a finding in a workpaper today. Could you produce the trail or just the tool's final answer?
Who signs off, and is it written down? Every letter agreed that a human has to own the final judgment. If your firm's answer to "who checked this" is "the AI checked itself," that's the gap every letter is trying to guard against.
Would your firm know if a citation or a fact was invented? KPMG, Deloitte and EY each found out the hard way, after publication, not before. A second-review step for anything AI drafted, even internal materials, is cheaper than a public retraction.
None of this waits on a formal PCAOB standard. The letters already told you what the guidance will probably ask for. Build the answer now, and the eventual requirement becomes a formality instead of a scramble.
Nexairi tracks PCAOB and audit-standard developments like this as they land. Subscribe to Nexairi Dispatch for the next update.
Sources
- PCAOB: PCAOB Seeks Stakeholder Input on Its Standard-Setting and Research Agendas
- Going Concern: Firms Have Weighed in on the PCAOB Turning an Eye to AI
- Big4News: PwC Tells PCAOB to Make AI in Audit Its "Foremost Priority"
- Bloomberg Tax: It's Time for the US Audit Board to Set Corporate AI Standards
- Big4News: The Hallucination Trap
- Grant Thornton: Comment on PCAOB Agenda Priorities
Related Articles on Nexairi
Free Assessment
Is your firm ready for AI?
A 5-minute governance check for CPA firms using ChatGPT, Copilot or AI accounting software. Get your score and your top gaps — free.
Jim Smart is the founder and editor in chief of Nexairi. A Business Intelligence Developer with experience building data systems for Verizon, U.S. Army operations, and enterprise finance teams, Jim spent years turning complex data into decisions that executives could act on — dashboards, forecasting models, and automation pipelines across telecom and government contracting. He founded Nexairi to apply that same clarity to AI: making emerging technology understandable and actionable for the operators, accountants, and business owners who need it most. Jim holds GenAI certifications from the University of South Florida Bellini College of AI and completed Springboard's Data Science Career Track.
