AI Ethics and Professional Standards

Imagine it’s late March, and your accounting practice is swamped with IRD provisional tax deadlines. You’re using Xero to generate a client’s tax estimate, and an AI tool suggests a deduction that seems too good to be true. You notice the AI output doesn’t align with the Income Tax Act 2007, but the tool is fast and convenient. Should you trust it? What if the client later disputes the advice, and you’re asked to explain your decision?

This scenario highlights the ethical and professional challenges of using AI in accounting. While AI can save time and improve efficiency, it must be used responsibly to maintain trust, comply with regulations, and uphold the standards set by CA ANZ (Chartered Accountants Australia and New Zealand).

All company names and scenarios used in this course are fictitious and created for illustration and training purposes only. Any resemblance to real businesses or organisations is coincidental.

1. CA ANZ Code of Ethics and AI

The CA ANZ Code of Ethics requires accountants to act with integrity, objectivity, and professional scepticism. When using AI, this means:

  • Always verifying AI outputs against known rules (e.g., the Tax Administration Act 1994) and client-specific information.
  • Avoiding over-reliance on AI for critical decisions, such as tax advice or financial reporting.
  • Maintaining independence by not allowing AI to influence professional judgment.

For example, if an AI tool flags a client’s GST return as error-free, you must still review it manually to ensure compliance with the GST Act 1985 and check for anomalies like unusually high expenses.

2. Professional Scepticism: Never Trust AI Unchecked

AI systems can make errors-especially when trained on incomplete or biased data. As an accountant, you must apply professional scepticism by:

  • Cross-checking AI outputs with manual calculations or other tools.
  • Questioning assumptions in AI-generated reports (e.g., why a tax liability was calculated differently).
  • Documenting your review process to show due diligence.

Example: An AI tool suggests a client can claim a $10,000 deduction for home office expenses. You must verify if the client meets the criteria in the Income Tax Act 2007 (e.g., working at least 20 hours per week at home) and check for consistency with their Xero records.

3. Client Confidentiality: What Data Can You Use?

AI tools often require access to sensitive client data. Under the Privacy Act 2020, you must:

  • Only use data that clients have explicitly consented to share.
  • Avoid storing or transmitting client data on unsecured AI platforms.
  • Ensure third-party AI providers comply with New Zealand privacy laws.

For instance, if you use an AI tool to analyse a client’s financials, you must confirm that the tool does not retain data beyond the agreed scope and that it’s encrypted during transmission.

Clients have the right to know when AI is used in their matters. Under the Tax Administration Act 1994, you must:

  • Disclose AI usage in writing, explaining its purpose and limitations.
  • Obtain consent before using AI for tasks like tax planning or compliance checks.
  • Be transparent about potential errors or biases in AI outputs.

Example: When preparing a client’s annual return using AI, you might say: “We used an AI tool to help calculate your provisional tax liability. While this speeds up the process, we’ve manually reviewed all figures to ensure accuracy.”

5. Bias in AI Outputs

AI systems can inadvertently reflect biases in their training data. For example:

  • A tool might under-estimate tax liabilities for clients in certain industries if its training data is skewed.
  • It could misclassify expenses (e.g., treating a business meal as private).

To mitigate this, always:

  • Test AI tools with diverse scenarios (e.g., different client types, industries, or tax situations).
  • Review outputs for fairness and alignment with the law.

6. Record Keeping

The Financial Reporting Act 2013 and Tax Administration Act 1994 require you to keep records of AI usage. This includes:

  • Documenting when and how AI was used (e.g., “AI tool X was used to generate this report on 15 April 2024”).
  • Storing AI-generated outputs alongside your manual work.
  • Retaining logs of AI decisions for audit purposes.

Example Prompts: Ready-to-Use Templates

Prompt 1: Verify AI tax advice
“I need to check if an AI-generated tax strategy complies with the Income Tax Act 2007. The tool suggests claiming a deduction for a client’s home office expenses. What are the legal requirements for this deduction in New Zealand?”

Prompt 2: Ensure client confidentiality
“How can I use an AI tool to analyse a client’s financials without violating the Privacy Act 2020? What steps should I take to secure their data?”

Prompt 3: Identify bias in AI outputs
“An AI tool calculated a client’s GST liability as $500 less than expected. What are the possible reasons for this discrepancy, and how can I investigate it?”

Prompt 4: Document AI usage
“What should I include in my records when using an AI tool to prepare a client’s tax return? Please provide an example of a compliance note.”

Common Pitfalls and Misconceptions

  • Assuming AI is always accurate: AI tools can make errors, especially with complex or non-standard cases. Always verify outputs manually.
  • Failing to disclose AI use: Clients have a right to know when AI is involved. Not disclosing this could breach the CA ANZ Code of Ethics.
  • Overlooking AML/CFT obligations: If using AI for client due diligence (e.g., screening for tax evasion), ensure it complies with the Anti-Money Laundering and Countering Financing of Terrorism Act 2009.

Try This: A Practical Exercise

Step 1: Use a free AI tool (e.g., Microsoft Copilot) to generate a short client communication template about using AI for tax advice. Example:
“We used AI to help prepare your tax return this year. While this improved efficiency, we manually reviewed all calculations to ensure accuracy.”

Step 2: Review the template for compliance with the Privacy Act 2020 and CA ANZ Code of Ethics. Ask:

  • Does it explain the purpose of AI use?
  • Does it mention manual verification?
  • Does it reassure the client about data security?

Step 3: Save the revised template for future use.

Key Takeaway

Using AI in accounting requires balancing efficiency with ethical responsibility. Always verify AI outputs, ensure client confidentiality, obtain informed consent, and document your work. By adhering to the CA ANZ Code of Ethics and New Zealand laws, you can harness AI’s benefits while maintaining trust and compliance.