FRC guidance explains when auditors can share AI-generated insights with clients
Insights from data tools can help companies, the regulator says, but presenting them as a showcase for advisory work risks the auditor's independence.
By Daily Oracles News Team2 min read
The Financial Reporting Council expects audit firms using AI and data analytics to turn up more observations about the companies they audit, and has set out when passing those findings on could compromise an auditor's independence.
The guidance, published on Monday, was written after auditors asked the regulator how to apply its Ethical Standard in situations that often involve technology. It is non-authoritative, and the FRC says it adds no new requirements to the existing ethical and independence rules.
AI can carry out new kinds of audit procedure that generate such insights, and it is also pushing firms to use older tools, such as data analytics, more widely.
Findings the audit already requires
Some communication is required anyway. The UK's auditing standards oblige auditors to report significant findings and weaknesses in internal controls to a company's board and management, and the FRC says that need not raise any independence question. One example in the document is an auditor using machine learning to spot purchases that, taken together, should have gone to a more senior manager for sign-off.
Insights beyond the audit
The harder cases are insights that go further. The guidance says these can amount to a non-audit service, or be seen as a favour if no market fee is charged for them.
It also warns of a self-interest threat when sharing an insight suggests the firm is looking for paid work later on. That is more likely if the insight covers a subject on which the firm offers advisory services, if colleagues from outside the audit team join the presentation, if the firm mentions that it can provide related non-audit services, or if the timing is close to the company buying such services. Insights drawn from emerging or proprietary technology that the client does not have can heighten the threat, the FRC says, because they may look like a demonstration aimed at winning more work.
Firms that give clients direct access to their tools face similar questions. Temporary, limited access to a process-mining tool to discuss a specific control weakness may be part of the audit. Continuing access that lets management monitor its own controls is less likely to be.
Audit committee approval
For public interest entities, a non-audit service of this kind is permitted only if the audit committee approves it after assessing the threats to independence, and it counts towards a cap on fees. The FRC says the question of whether independence would be compromised will often decide whether an insight can be shared, rather than a simple test of whether it counts as a non-audit service.
The document follows the FRC's earlier guidance on generative and agentic AI tools in audit work and on possible uses of AI in audit.




