The ATO as an AI user: what it means for your compliance profile
The ATO is one of the most sophisticated users of AI in the Australian public sector. Its compliance programs use machine learning to identify unusual patterns in tax data, to risk-score individual and business taxpayers for audit attention, and to detect potential fraud and evasion. The specific algorithms are not public, but the ATO has disclosed that its data matching programs, risk detection systems, and compliance analytics use AI and machine learning techniques.
Understanding that your tax returns are assessed against AI-generated risk profiles has practical implications. Returns that are statistically anomalous — whether because of legitimate but unusual deductions, industry-specific characteristics, or genuine errors — may receive increased scrutiny. The ATO's AI systems are trained on historical compliance data, which means they reflect historical patterns of what compliant and non-compliant returns look like. Businesses whose tax profile deviates significantly from industry norms for reasons that are legitimate and explainable should maintain documentation that supports that explanation.
AI-assisted tax compliance: getting it right
Many businesses now use AI tools to assist with tax preparation — software that suggests deductions, categorises expenses, or calculates GST obligations. The ATO's position is clear: the taxpayer is responsible for the accuracy of their tax returns regardless of the method used to prepare them. An AI tool that generates an incorrect deduction claim does not transfer responsibility for that claim from the taxpayer to the AI provider. Due diligence on the accuracy of AI-assisted tax calculations is a taxpayer obligation.
Specific areas of ATO focus for AI-assisted tax compliance: deduction classification — AI tools that systematically misclassify expenses (for example, categorising capital expenditure as deductible operating expense) create liability for the taxpayer; GST calculations — AI invoicing and accounting tools must correctly apply GST rules including the complex rules for mixed supplies and financial services; and R&D tax incentive claims — AI-generated documentation for R&D claims must meet the ATO's substantiation requirements, including specific documentation of eligible activities and expenditure.
Transfer pricing and AI allocation methodologies
Multinational enterprises using AI to allocate costs, revenues, or profits between related parties in different jurisdictions face specific transfer pricing documentation obligations. The ATO's transfer pricing rules require that cross-border related-party transactions be priced on arm's length terms, with documentation that demonstrates this. Where AI is used to generate the allocation methodology or the pricing model, the documentation must explain the AI's methodology clearly enough for the ATO to assess whether the outcome is arm's length. AI-generated transfer pricing that cannot be explained and audited creates documentation risk.
