ICAEW Publishes AI Assurance Framework for Audit and Accounting Firms

What happened

The Institute of Chartered Accountants in England and Wales (ICAEW) published its long-anticipated AI Assurance Framework on 17 April 2026, providing detailed practical guidance for accounting and audit firms deploying artificial intelligence in their workflows. The framework addresses the full lifecycle of AI adoption — from initial risk assessment and vendor due diligence through to ongoing monitoring, human oversight requirements, and client disclosure obligations.

The framework is structured around five pillars: governance and accountability, data integrity, model validation, transparency and explainability, and professional judgement. It makes clear that while AI tools can enhance efficiency in areas such as transaction classification, anomaly detection, and document review, the ultimate responsibility for professional conclusions rests with qualified individuals. Firms are expected to maintain auditable records of how AI outputs contributed to any engagement deliverable.

Notably, the ICAEW framework aligns with the Financial Reporting Council's (FRC) existing expectations around audit quality and explicitly references the EU AI Act's high-risk classification criteria as a useful reference point, even though the UK is not directly bound by that legislation. The guidance also addresses the use of large language models (LLMs) such as GPT-4.1, Claude, and Gemini, warning that firms must validate outputs for hallucination risk and ensure client data is not inadvertently shared with third-party model providers.

The ICAEW stressed that the framework is not legally binding but represents best practice that regulators — including the FRC during audit inspections — are likely to reference when assessing firm conduct. A consultation on supplementary technical notes, including sector-specific guidance for tax advisory and insolvency work, will open in Q3 2026.

What this means for your business

For UK accounting and audit firms, this framework effectively sets the professional standard against which AI usage will be judged. Even though it is described as guidance rather than regulation, the ICAEW's explicit statement that the FRC will reference it during inspections gives it significant practical weight. Firms that have been experimenting with AI tools — whether for audit sampling, management accounts preparation, or advisory research — now have a clear checklist against which to evaluate their processes. Those without documented AI governance policies should treat this as an urgent priority.

The framework's emphasis on vendor due diligence and data handling is particularly relevant given the rapid adoption of cloud-based AI tools across the profession. Firms using Microsoft Copilot, OpenAI's API, or Anthropic's Claude for client work will need to demonstrate that they understand where client data flows, what retention policies apply, and whether outputs have been independently validated. The guidance on LLM hallucination risk also underscores the need for human review processes that go beyond cursory checks.

More broadly, the ICAEW's move signals that UK professional bodies are moving from general AI principles to specific, actionable standards — a trend that law firms, consultancies, and financial advisers should watch closely. The SRA's generative AI guidance published earlier this month, combined with the FCA's model risk consultation, suggests a coordinated tightening of expectations across the professions. Firms that invest now in robust AI governance will be better positioned both for regulatory scrutiny and for client confidence as AI becomes embedded in routine professional work.