Artificial intelligence is on the Australian Taxation Office’s radar, with a senior official warning those who rely on AI tools to generate information to tread carefully.
ATO second commissioner Jeremy Hirschhorn said he is concerned people looking to comply with the law will rely on tools that produce unreliable information — especially if those AI responses use information sourced overseas.
“For those who have experimented with ChatGPT and other AI engines, you will have come across the problem that they can generate absolutely compelling, plausible but completely incorrect responses,” Hirschhorn told an international tax administration conference.
“This is exacerbated in the area of tax, where fine differences can matter within jurisdictions. But also differences between jurisdictions’ approaches mean that an AI response which is completely correct in relation to one country will be completely incorrect in another.”
With the ATO midway through a digitisation project that aims to make it easier for taxpayers to manage tax affairs, Hirschhorn said he was worried about other problems created by the digital revolution.
“The ATO currently holds about 50 petabytes of actively used data and processes about 20 billion transactions each year,” Hirschhorn said.
“On any given day, our systems block an average of approximately 89,000 malicious connections. This is even higher during our peak individuals lodgement period — ‘tax time’.
“We are actively hardening our current systems against the rise of cyber-enabled fraud at scale (such as identity and information theft) but also embedding fraud prevention measures into systems as part of the initial design process for whatever the future may bring.”
The ATO reports more than $30 billion in liabilities have been raised against large corporations and multinationals since the Tax Avoidance Taskforce was established in 2016.
Hirschhorn said the avoidance taskforce accounted for more than half of the total liabilities — $18.7 billion — as the ATO ramped up enforcement activities related to big companies attempting to dodge tax payments.
The taskforce, which is set to have 3,000 members, has meant the ATO can switch from better-known, traditional tax audit methodology to proactive engagement, Hirschhorn said.
In part, this is because the ATO methodology now embeds what it calls a “justified trust” approach. There are companies that pay the right amount of tax, and the revenue authority can trust its dealings with them.
“Over half of the top 100 have attained an overall high assurance rating, meaning they have provided the ATO with objective evidence that they are paying the right amount of tax,” Hirschhorn said.
“These programs were primarily designed to encourage increased compliance by large businesses through increased coverage and detection of incorrect tax positions, and long-term behavioural change due to the knowledge they will be reviewed.”
Hirschhorn told the conference the ATO’s proactive approach has resulted in voluntary compliance.
“Thanks to this prevention-first approach, voluntary compliance in the large market segment is also the highest it has ever been, with our most recent estimate being 93% for 2019–20, which is up from around 90% in 2014–15,” Hirschhorn said.
“Overall performance is about 96% after compliance activity.”
The ATO attributes a part of its success to the work of international bodies.
“Our work in Australia could also not have been possible without the significant reforms introduced by the OECD to address digitisation of the economy over the past 10 years and successive Australian governments’ commitment to their implementation,” Hirschhorn said.
This article was first published by The Mandarin.
Tom Ravlic is a journalist and author with a deep expertise in accounting, corporate governance and regulatory affairs. He has worked as an accounting academic, policy adviser for one of Australia’s three major accounting bodies, and consulted on the development of accounting curriculum for universities and professional associations.
