CPA Australia suggests tweaks to TPB’s draft AI guidance

“While AI, particularly generative AI, is seen by many as having a revolutionary impact on the professional services sector in terms of productivity, and simultaneously potentially an existential threat to those working in that sector, the guidance does not get drawn into such debates,” the submission read.

“Instead, it appropriately focuses on the obligations and responsibilities of those using AI when providing tax or BAS agent services.”

While CPA Australia said it supported the proposed guidance, it offered some technical amendments. Firstly, it suggested the guidance should more strongly emphasise the fact that AI may not always be accurate or factually correct.

The TPB guidance urged tax practitioners to verify and review AI generated content for accuracy, and establish processes to understand and contest AI outputs.

CPA Australia also asked for more clarity regarding the TPB’s definition of independent verification, with respect to its reasonable care requirements.

“Whether a tax practitioner has taken reasonable care in a given situation will depend on an examination of all the circumstances, including but not limited to … whether AI tools have been used for tasks which should be independently verified,” TPB documentation read.

CPA said it was unclear whether the TPB meant outputs should be ‘independently verified’ by the tax agent themself or by somebody else, especially given that the previous dot point already stipulated the TPB would consider “whether the tax practitioner checks or reviews the output of the AI tool before purporting to rely on it.”

It also noted there was a “lack of clarity” regarding CPA’s definition of “reasonable care,” which warranted consideration.

The industry body added that paragraphs 27 and 28, which concerned “other code obligations,” were “not guidance” and should be amended to be more relevant to AI, or removed altogether.

CPA Australia also praised the draft guidance for its principles-based approach, which it said would likely continue to stand as AI tools rapidly evolved.

“As we have seen already, the AI landscape is fast evolving with new and emerging technological advancements,” its submission read. 

“However, by retaining a principles-based approach in the guidance rather than being prescriptive about the type of AI that may be used, the guidance should remain relevant and current in the face of technological changes.”