Automated systems and AI in Australian law: what has been enacted and decided

What Australian courts and the Commonwealth Parliament have actually decided and enacted about automated systems and AI as at October 2026: AI inventorship after Commissioner of Patents v Thaler, computer-generated letters and administrative decisions after Pintarich, the automated-decision disclosures in APP 1.7 to 1.9 from 10 December 2026, and the sexual deepfake offence in s 474.17A of the Criminal Code.

Learning level
Core Doctrine
Jurisdictions
au-commonwealth
Subjects
intellectual-property-and-technology-law
Topics
emerging-technology-regulation

Learning outcomes

  • Explain why, after Commissioner of Patents v Thaler, an AI system cannot be named as the inventor under s 15(1) of the Patents Act 1990 (Cth), and identify what the Full Court and the High Court left undecided.
  • Apply the majority test in Pintarich v Deputy Commissioner of Taxation to decide whether a computer-generated communication records a decision, and distinguish the dissent.
  • Determine whether APP 1.7 to 1.9 of the Privacy Act 1988 (Cth) will require an entity’s privacy policy to describe its automated decisions, and from when.
  • Identify the elements, exceptions and aggravated form of the offence in s 474.17A of the Criminal Code (Cth) as they apply to digitally created sexual material.

Many questions about artificial intelligence and automated systems are still open in Australian law, but some have now been answered by a court or by Parliament. This article states only those answers, all of them Commonwealth law, as they stood in October 2026: who can be an inventor, when a computer-generated letter is a decision, what a privacy policy must say about automated decisions, and when sharing a digitally created sexual image is a crime. Faced with a technology problem, a student must first ask whether the question is one an enacted provision or a decided case actually answers, because where one does, the answer is the ordinary law applied to new facts, and the genuine uncertainty lies in what it leaves unanswered.

An AI system cannot be named as the inventor

A patent for an invention may only be granted to a person who is the inventor, or who derives entitlement from the inventor by assignment, title or representation (Patents Act 1990 (Cth) s 15(1)).1 Dr Stephen Thaler applied for a patent naming as inventor “DABUS”, an artificial intelligence system, and stating that the invention had been autonomously generated by it. The Deputy Commissioner of Patents treated the application as lapsed because it did not name a natural person as inventor, but a single judge of the Federal Court set that aside.

In Commissioner of Patents v Thaler [2022] FCAFC 62, a Full Court of five judges (Allsop CJ, Nicholas, Yates, Moshinsky and Burley JJ) allowed the Commissioner’s appeal.2 Their Honours held that the reference in s 15(1)(a) to “a person who is an inventor” emphasises, in context, that the inventor is a natural person, and that each of the other paragraphs of s 15(1) derives entitlement from that inventor (at [106]–[109]). Naming DABUS as the inventor therefore did not comply with the Patents Regulations (at [117]). The Court was careful about what it was not deciding: it rejected the premise that no invention devised by an artificial intelligence system could be patented, and said that whether this application had a human inventor had not been explored and remained undecided (at [121]). It also said that the wider policy questions require consideration with “some urgency” (at [120]).

The High Court refused Dr Thaler special leave to appeal on 11 November 2022, saying that the case was “not the appropriate vehicle to consider the questions of principle” he sought to raise.3 The Full Court’s decision therefore stands, but the High Court has not itself decided the question.

A computer-generated letter is not necessarily a decision

In Pintarich v Deputy Commissioner of Taxation [2018] FCAFC 79, a tax officer had keyed information into a computer-based template bulk-issue letter, and the letter sent to the taxpayer appeared to accept a lump sum in full discharge of his tax debt and general interest charge.4 The taxpayer paid, but later statements of account showed interest still owing, and in May 2016 the Deputy Commissioner granted only partial remission of it; the taxpayer argued that the December 2014 letter had already decided to remit the interest (at [102]–[103], [116], [120]). Moshinsky and Derrington JJ held that, for there to be a decision to remit the interest, there needs to be both a mental process of reaching a conclusion and an objective manifestation of it (at [140]). Because the officer had not reached any conclusion on the application to remit interest, no decision had been made, even if the letter was read as the taxpayer contended (at [145], [153]). The majority described the circumstances as “quite unusual” (at [152]), and accepted that the general statement of what making a decision involves “may not be applicable in relation to all issues” (at [148]). In eSafety Commissioner v Baumgarten [2026] FCAFC 12, a later Full Court held that an eSafety officer who sent a “complaint alert” to the platform X had made a decision, because she went through a mental process and translated it into action (Mortimer CJ and Beach J at [192]); Horan J applied the Pintarich formulation (at [221]) and noted that Pintarich arose in a different context, where the taxpayer was trying to prevent a later, less favourable exercise of the same power (at [223]).5

Kerr J dissented. His Honour observed that the expectation that a decision involves human mental processes “is being challenged by automated ‘intelligent’ decision making systems”, and that the legal conception of a decision cannot be static (at [46], [49]). Those observations are a dissent and not the law the case laid down. As a Full Federal Court decision, Pintarich binds single judges of the Federal Court and is persuasive elsewhere; it concerned a template letter generated from an officer’s input, not a decision made entirely by a computer program.

Privacy policies must describe automated decisions from 10 December 2026

The Privacy and Other Legislation Amendment Act 2024 (Cth) adds three new subclauses to Australian Privacy Principle 1 in Schedule 1 to the Privacy Act 1988 (Cth), commencing on 10 December 2026.6 Under APP 1.7, an APP entity’s privacy policy must contain the information in APP 1.8 if:

  1. the entity has arranged for a computer program to make, or to do a thing substantially and directly related to making, a decision;
  2. the decision could reasonably be expected to significantly affect the rights or interests of an individual; and
  3. personal information about the individual is used in the operation of the computer program to make the decision or do that thing.

The policy must then describe the kinds of personal information used, the kinds of decisions made solely by the operation of such programs, and the kinds of decisions for which a program does a thing substantially and directly related to making them (APP 1.8). A decision includes refusing or failing to make one, and the examples given are decisions under an Act or legislative instrument to grant or refuse a benefit, decisions affecting rights under a contract, and decisions affecting access to a significant service or support (APP 1.9). The amendment applies to decisions made after commencement, even if the arrangement or the personal information predates it. Read on its terms, APP 1.7 is a disclosure obligation about what the privacy policy must contain; it does not itself set rules for how the automated decision is made.

Sharing sexual deepfakes is a Commonwealth offence

Since 3 September 2024, s 474.17A of the Criminal Code (Cth) has made it an offence to use a carriage service to transmit material of another person who is, or appears to be, 18 or older, where the material depicts or appears to depict that person engaging in a sexual pose or sexual activity, or depicts sexual organs, the anal region or a female person’s breasts, and the transmitter knows the person does not consent or is reckless as to consent.7 The maximum penalty is imprisonment for 6 years. It is irrelevant whether the material is unaltered or has been created or altered in any way using technology, and the note to s 474.17A(2) gives “deepfakes” as examples. Recklessness includes not giving any thought to whether the person consents, and “transmit” includes make available, publish and distribute.

The offence does not apply where transmission is necessary for, or of assistance in, enforcing or investigating a law, is for the purposes of proceedings in a court or tribunal, is for a genuine medical or scientific purpose, or is one a reasonable person would consider acceptable having regard to listed matters; the defendant bears an evidential burden on those exceptions (s 474.17A(3)). An aggravated offence with a maximum of 7 years applies where the person who transmitted the material was also responsible for creating or altering it (s 474.17AA(5)).

Applying this in a problem question

  1. Identify the precise legal question the technology raises, such as inventorship, whether a decision was made, a privacy-policy obligation or a criminal transmission.
  2. Check whether an enacted provision or a decided case answers that question, and note its commencement date and the court that decided it.
  3. Apply the answer as ordinary law: name a natural person as inventor, ask whether a decision-maker actually reached a conclusion, test the three APP 1.7 conditions, or work through the elements and exceptions of s 474.17A.
  4. Distinguish a holding from a dissent or an open question, such as whether an invention devised with an AI system has a human inventor.
  5. State plainly which parts of the problem remain unanswered by the enacted and decided law.

Pop quiz

5 quick questions on this article, the authorities it cites and the articles it links to.

  • About 3 minutes, and no time limit.
  • You can only go forwards: each answer locks when you submit it.
  • After each question you see the right answer, why, and where to read more.
  • Free, and no account needed. Log in or create a free account to keep your scores.