The Australian Taxation Office has used artificial intelligence on an industrial scale to identify more than $530 million of unpaid tax bills and stop $2.5 billion being fraudulently claimed.
ATO deputy commissioner Marek Rucinski said the agency’s deep learning models had helped his staff identify $295 million in superannuation guarantee underpayments, and natural language models had scoured leaked documents, such as the Panama Papers, to detect $242 million owed by tax evaders since 2018.