David Rotfleisch on if the CRA Is Using Flawed AI To Audit You: Lessons for Canadian Taxpayers From Australia’s Robodebt Scandal

Overview

The Canada Revenue Agency (CRA) is rapidly expanding its use of artificial intelligence (AI), machine learning, and advanced data analytics to detect potential tax non-compliance. By analyzing vast amounts of data from tax returns, financial institutions, payroll records, cryptocurrency platforms, and international sources, the CRA can more efficiently flag files for review.

While these tools can strengthen tax enforcement, they also raise serious concerns about accuracy, transparency, procedural fairness, and taxpayer rights. Australia’s infamous Robodebt scandal offers a stark warning for Canadian taxpayers.

In Robodebt, an automated debt recovery system issued hundreds of thousands of incorrect debt notices based on flawed income-averaging assumptions and inadequate human oversight. The program was later ruled unlawful, forcing the Australian government to repay millions and pay substantial settlements. Former U.S. National Taxpayer Advocate Nina Olson has cited Robodebt as a cautionary tale for tax authorities relying heavily on automated systems.

As the CRA invests more in AI-driven compliance, Canadian taxpayers must understand both the benefits and the risks of technology-powered tax enforcement.

What Was Australia’s Robodebt Program?

Robodebt was designed to recover alleged overpayments of welfare benefits. The system automatically matched data and used income averaging to calculate debts. Because many recipients had irregular or seasonal income, the methodology produced widespread errors.

Hundreds of thousands of Australians received inaccurate debt notices. Many lacked the resources or information to challenge them effectively. Courts ultimately found key elements of the program unlawful. The scandal highlighted the dangers of over-reliance on automated decision-making without proper safeguards.

How the CRA Uses Artificial Intelligence

The CRA openly uses advanced analytics and machine learning to identify higher-risk files in areas such as:

• Cryptocurrency tax compliance
• Offshore assets and foreign income reporting
• Unreported business income
• GST/HST compliance
• Payroll and aggressive tax planning
• High-net-worth taxpayer reviews

Third-party data from banks, payment processors, crypto exchanges, and international agreements further powers these systems. While this improves efficiency, it also increases the risk that incomplete or misleading data can trigger unnecessary CRA audits.

Important distinction: AI tools flag potential issues; they do not determine actual tax violations. That requires human review and evidence.

The Risks of AI-Driven Tax Audits

Automated systems are only as good as their data, assumptions, and algorithms. Common problems include:

• False positives: Legitimate transactions or business patterns that differ from statistical averages can trigger audits.
• Systematic errors affecting thousands of taxpayers.
• Significant time, stress, and professional fees for taxpayers who ultimately prove full compliance.

A small error rate applied across millions of returns can still create major problems, as Robodebt clearly demonstrated.

Why Human Oversight Is Essential

Tax law is complex and highly fact-specific. AI can spot patterns but cannot understand context, intent, or supporting documentation the way an experienced auditor can.

The key lesson from Robodebt is clear: Technology should support decision-making, not replace it. Proper human review before issuing assessments or initiating enforcement actions is critical to protecting taxpayer rights.

Procedural Fairness and Taxpayer Rights

Canadian taxpayers have the right to:

• Understand the basis of any CRA action
• Provide supporting documentation
• Challenge assumptions and assessments
• Appeal where necessary

As AI use grows, courts will likely place greater emphasis on the transparency and explainability of automated decisions. Taxpayers should never be left unable to meaningfully respond to CRA concerns generated by algorithms.

Pro Tax Tips

• Maintain complete, organized records for all income, deductions, credits, and transactions—especially for crypto, foreign assets, self-employment, and shareholder loans.
• If selected for a CRA audit, do not assume the CRA is correct. Carefully review their assumptions and evidence.
• Seek advice from an experienced tax lawyer in Toronto early. Early intervention can resolve misunderstandings, prevent positions from becoming entrenched, and dramatically improve outcomes.

Frequently Asked Questions

Does the CRA use artificial intelligence?

The Canada Revenue Agency (CRA) uses advanced analytics, machine learning, and automated risk assessment tools to help identify tax compliance risks, detect potential errors or inconsistencies, and select files for further review. These technologies allow the CRA to analyze large volumes of tax data more efficiently and focus its audit and enforcement resources on higher-risk cases.

Can AI decide if I will be audited?

AI and automated systems may help identify taxpayers or tax returns that warrant further scrutiny, but they are generally used as decision-support tools rather than final decision-makers. Human auditors and CRA officials are expected to exercise professional judgment when determining whether to proceed with an audit, review, reassessment, or other enforcement action.

What is a false positive in this context?

A false positive occurs when a taxpayer is incorrectly identified as high-risk or non-compliant even though they have complied with their tax obligations. This can happen when automated systems rely on incomplete, inaccurate, or misleading data, or when legitimate transactions resemble patterns associated with tax non-compliance. False positives can result in unnecessary audits, reviews, or requests for additional information.

What should I do if I receive a CRA audit notice that I believe is based on incorrect information?

Remain calm and carefully review the audit notice and any concerns identified by the CRA. Gather all relevant records, receipts, financial statements, and supporting documentation that may address the issues raised. Respond to CRA requests within the required deadlines and keep copies of all correspondence. If you believe the audit is based on inaccurate information, a misunderstanding of the facts, or an improper assessment of risk, you should consult an experienced Canadian tax lawyer as soon as possible to protect your rights and help you navigate the audit process.

Can I challenge a CRA decision that was influenced by automated systems or AI?

Taxpayers generally have the right to request explanations, provide additional information, object to reassessments, and pursue appeals through the CRA’s objection process and, where necessary, the courts. If you believe an automated system contributed to an incorrect outcome, obtaining legal advice can help you understand your options and build a strong response.

Key Takeaway

Artificial intelligence offers powerful tools for tax administration, but Australia’s Robodebt scandal shows the serious harm that can result from over-reliance on automation without adequate safeguards.

As the CRA expands its use of AI and data analytics, meaningful human oversight, transparency, and respect for taxpayer rights remain essential. Canadian taxpayers should support efficient tax administration while staying vigilant in protecting their rights during any CRA interaction.

Disclaimer: This article provides general information only and is accurate as of the date of posting. It is not legal or tax advice. Every situation is unique. For advice specific to your circumstances, please contact an experienced Canadian tax lawyer.