A CCAB Draft Case Study Says Client Authorization Is Still Required Even When Client Data Is Processed to Create Synthetic Data

In June 2026 the Consultative Committee of Accountancy Bodies, CCAB, published draft case studies on the ethical use of AI, marked as draft guidance for consultation. There are six. They are "fictional scenarios intended to highlight potential ethical dilemmas accountants may face when engaging with AI," and the foreword says they "were developed with the assistance of AI tools and have been quality assured by members of the CCAB Ethics Group." The disclaimer says they are not intended as legal or professional advice. Parts of two of them follow, the third case study most of all.

A Model Trained on Synthetic Client Data

In the third case study the reader is a partner responsible for innovation and technology at a large global accounting firm that is developing its own large language model. The technology team proposes a synthetic dataset derived from actual client data: client financial data, transaction records, contracts and other sensitive information would be used to generate data that "maintains the statistical properties and patterns of the original data but doesn't contain actual client information." The team argues that since the actual client data is not directly used to train the model, client confidentiality is maintained.

Concerns are raised. The synthetic data might still retain "patterns unique to specific clients that could potentially be extracted or inferred by sophisticated users." The firm's engagement letters and terms of business "don't explicitly mention using client data, even in anonymised or synthetic form, for training AI systems." There is uncertainty, it says, about whether clients from regulated industries would have restrictions of their own. And the technology team acknowledges "a small but non-zero risk" that actual client data could occasionally leak through into the synthetic dataset.

The Guidance, and the Recommendations

The case study says recent guidance from the Financial Reporting Council "has clarified several important points relevant to this situation." Among them: "The IESBA Code requires authorisation from the audited entity before any confidential information is used in the development of technology." And "Even when client data is processed to create derived or synthetic data, authorisation from the client is still required." The reason it gives is that the first step, processing the original client data, "constitutes using confidential information for the development of technology, albeit as a preliminary step." Creating synthetic data from patterns observed by human expertise over time, "without directly processing client data," would be permitted under the IESBA provisions, according to the case study's account of that guidance.

The step that needs the client's permission comes before any model is trained: it is the moment the original data is processed.

The case study's recommendations include a formal opt-in process "for existing clients rather than assuming implied consent," updated standard terms for new clients, an ongoing monitoring program "to regularly test synthetic data for potential client information leakage," and documenting "these decisions and processes thoroughly to demonstrate due care and professional judgement in alignment with both the letter and spirit of the ethical standards."

The Audit Platform in the Second Case

The second case study is a mid-sized firm using an AI audit platform on an engagement. Among the issues listed, the system's extraction tools captured incomplete data for parts of the client's own system, the documentation of procedures "has become less detailed as team members rely on the AI system's automated workpapers," and the client's data is processed through the system's cloud servers, "raising questions about data security and confidentiality that weren't fully addressed in the engagement letter." Under confidentiality, the case study asks whether "adequate safeguards and client permissions" were secured for processing and storing sensitive client financial data "on the third-party AI system's cloud servers."