The Financial Reporting Council (FRC) published guidance on 16 November on how audit firms should protect their independence when they share insights from AI and other technology with the companies they audit.
The guidance is non-authoritative: it explains how existing ethical and independence requirements apply and sets no additional ones. The FRC expects the volume of technology-derived insights to rise.
Mark Babington, Executive Director of Regulatory Standards at the FRC, said the use of AI by auditors can improve audit work and the insights given to audited entities. “However, auditor independence is and must remain at the heart of audits providing value and serving the public interest,” he said.
The regulator said it wrote the guidance after auditors asked for help applying its Ethical Standard in specific situations, often involving technology.
It covers two things: passing insights to management or to those charged with governance, and giving them access to technology so they can generate insights themselves.
The FRC’s position is that technology used appropriately can improve audit quality, for example by identifying risks of material misstatement or ways to strengthen internal controls.
The risk, it said, arises where the work behind an insight is not part of the audit of the financial statements.
The full text of the guidance sets out four threats. These are management, self-review, self-interest and familiarity.
On self-interest, the guidance lists factors including the topic of the insight and the source. Where an insight comes from emerging or proprietary technology that the audited entity may lack, the threat may be heightened. The reason given is that there may be a perception that the auditor is demonstrating the technology as a “basis for winning future work”.
The topic matters too.
An insight related to the audit or financial reporting process is treated differently from one on unrelated subjects.
Regularly offering insights on unrelated topics may lead an objective, reasonable and informed third party (ORITP, the FRC’s test for how a situation would look to an outsider) to see the auditor as too close to management.
Not all communication needs this analysis.
Where an auditing standard requires the auditor to tell those charged with governance about a matter, such as significant findings or deficiencies in internal control, the guidance says it is not a non-audit service and poses no threat to independence.
It adds that auditors are not expected to document a formal consideration for every insight, and formal review may often not be needed for insights derived from work done to support the audit opinion.
The management threat covers situations where an auditor’s insights lead to it effectively playing a part in running the client. The guidance lists opaque technology, a high volume of insights and the risk of automation bias as features that may heighten it.
Automation bias means management accepting an output without critical evaluation.
On self-review, the guidance warns that an AI model used to audit information shaped by earlier outputs of a similar model may be biased towards corroboration. It says this does not necessarily preclude using the model.
It also says insights in general thought leadership pieces fall outside its scope.
The guidance builds on earlier FRC guidance on generative and agentic AI tools in audit engagements and on potential use cases of AI in audit.
It ends with two illustrative examples. In the first, an audit team on a public interest entity uses an AI tool on revenue recognition and then does a small amount of separate work to produce customer insights for management.
No fee is charged, and the guidance concludes the communication would be a permitted non-audit service, subject to audit committee approval.






