澳大利亚信息专员办公室《APP 1.7—1.9 透明度义务配套资源——人工智能与自动化决策使用的透明度》（2026年9月30日发布，2026年12月10日起适用）

September 2026 Page 1 APP 1.7-1.9 Transparency Obligation Fact Sheet This fact sheet is intended to help APP entities implement new Australian Privacy Principle (APP) 1 transparency obligations under APP 1.7-1.9, with respect to the usage of computer programs to make or inform decisions (transparency obligation). Changes to privacy obligations for APP entities come into effect on 10 December 2026. This fact sheet does not cover the entirety of an APP entity’s privacy obligations and should be read in conjunction with the Privacy Act 1988 (Cth), Australian Privacy Principles, OAIC’s APP 1 Guidelines, privacy guidance on APP 1.7-1.9 and other OAIC resources referred to in this fact sheet. Government agencies should also refer to the supplementary fact sheet on the transparency obligation in the context of Freedom of Information and other transparency considerations. Please refer to the end of this fact sheet for practical examples about the application of the transparency obligation. When does the transparency obligation apply? An APP entity is in scope of the transparency obligation where:  it has arranged for a computer program to make, or do a thing that is substantially and directly related to making a decision; and  the decision could reasonably be expected to significantly affect the rights or interests of an individual; and  personal information about the individual is used in the operation of the computer program to make the decision or do the thing that is substantially and directly related to making the decision. APP entities in doubt of whether their use of a computer program falls within the scope of the transparency obligation should take a cautious approach and include information in their APP privacy policy necessary to comply with APP 1.8 requirements. Have you arranged for a computer program to inform or make a decision? A computer program includes:  pre-programmed rule-based processes  artificial intelligence and machine learning processes  software, apps or word-processing tools  generative AI used to generate text, images, videos, code or synthesis, including chatbots. A decision includes: • making a decision – choosing to take an action or not take an action September 2026 Page 2 • doing something – including deliberately doing something or failing/refusing to do something • a decision that affects a person – where the outcome could reasonably be expected to benefit or disadvantage that individual. Refer to Note 1 and Example 1 for additional information about the meaning of a decision. A decision may be within scope of the transparency obligation even where a computer program output does not replace the entire decision -making process or is subject to human review. Examples of circumstances where an APP entity may have ‘arranged for’ a computer program to inform or make a decision include (See Example 2):  use of an in-house or proprietary software capability  procurement of a third-party computer program  configuration or customisation of existing off-the-shelf or packaged software  authorisation of a computer program to make a determinative decision  reliance on advisory outputs from a computer program. Is the computer program substantially and directly related to making a decision? Substantially means where it is a key factor in facilitating the human’s decision making; and Directly means where the thing has a direct connection with making the decision. A computer program may be substantially and directly related to a decision if it:  advises of an appropriate outcome of a process  determines the outcome of a process  in any other way influences the outcome of a process. A computer program may be substantially and directly related to making a decision regardless of whether the nature of the output is advisory or determinative. In this regard, the OAIC considers machine learning or generative AI outputs used to make decisions that significantly affect the rights or interests of individuals would generally fall within scope of the transparency obligation unless subject to extensive human oversight and control. Example 3 discusses the application of substantially and directly related to making a decision. Does the decision significantly affect the rights or interests of an individual? Significantly means the impact of the decision or thing must be more than trivial and must have the potential to considerably influence the circumstances or outcomes for the individual concerned. A right is a moral or legal entitlement to have or be able to do something. An interest refers to a concern, benefit, stake or claim relating to something. September 2026 Page 3 Vulnerability is relevant to this limb of the transparency obligation, to the extent that a decision may have a greater impact on persons experiencing vulnerability, as opposed to the general population. Examples 4 and 5 provide further insight into the application of the significantly affect the rights or interests of an individual limb of the transparency obligation. What information must APP entities subject to the transparency obligation include in their APP 1 privacy policy? APP entities within the scope of the transparency obligation are required to include the following information in their privacy policy:  the kinds of personal information used in the operation of such computer programs  the kinds of such decisions made solely by the operation of such computer programs  the kinds of such decisions for which a thing, that is substantially and directly related to making the decision, is done by the operation of such computer programs.1 The OAIC generally expects that APP 1 transparency obligations typically remain with the APP entity who is using personal information to make a decision or do a thing that significantly affects the rights or interests of an individual. Providers of third-party software or enterprise solutions to APP entities making decisions that significantly affect the rights or interests of individuals should provide clear, high-level information about how their software can be used to make decisions to enable the APP entity to include appropriate information in their APP 1 privacy policies. For example, where an APP entity has procured a third-party computer program with machine learning capabilities to provide predictive analytics or personalisation when making a decision which significantly affects the rights or interests of an individual, the APP entity should disclose how personal information was processed by the computer program based on the information provided by their vendor. How do APP entities treat commercial-in-confidence information under the transparency obligation? Commercial-in-confidence information about computer programs is explicitly excluded from APP 1.7 requirements.2 The following information is not in scope of the transparency obligation:  commercially sensitive information  trade secrets. Information about an APP entity’s use of a computer program to make significant decisions that results in exposure to ridicule, embarrassment or public criticism does not fall within the commercial- in-confidence exclusion. 1 APP 1.8 2 Paragraph 38 of the Privacy and Other Legislation Amendment Act 2024 (Cth) Addendum to the Explanatory Memorandum September 2026 Page 4 The following factors may assist in deciding whether information is commercial -in-confidence: - whether the information about the kinds of personal information used or kinds of decisions made are unique to the APP entity, and if they were known to others, the extent to which it would detract from the information’s intrinsic commercial value - whether the information confers a competitive advantage on the APP entity — for example, if it lowers the cost of production or allows access to markets not available to competitors - whether a genuine ‘arm’s-length’ buyer would be prepared to pay to obtain that information. How much technical information is required to comply with the transparency obligation? APP entities should consider the degree of technical information they include in their privacy policy in the context of their individual activities and circumstances. APP entities should:  ensure individuals can understand how their personal information is handled  avoid overwhelming individuals with unnecessary amounts of granular technical detail  structure information to enable consumers to request further information, where required  tailor information to be specific the APP entity’s activities and circumstances  frame information in a way that allows the information and the decision to be reviewed. APP entities should consider the degree of technical information they include in their privacy policy in the context of their individual activities and circumstances. APP entities may group information about the categories of decisions and types of personal information in their privacy policy to comply with the transparency obligation. However, the level of detail provided about these processes must be meaningful to a reasonable person. APP entities may use a range of personal information and different computer programs to process a very large number of automated processes and decisions. These may be integrated together or performed in rapid succession to render an output that is ultimately used to make a decision. Where an APP entity chooses to group information about categories of decisions and types of personal information used to inform a decision, this should be considered with regard to the sensitivity of the information relied upon by the computer program. Where more sensitive types of information (e.g. health information or biometric templates) are used, this should be clear. Examples 6 and 7 provide illustrations on the extent of technical information required under the transparency obligation. September 2026 Page 5 Appendices Appendix 1: Notes on the APP 1.7-1.9 Transparency Obligation Note 1 – Decisions considered in scope of the transparency obligation The OAIC outlines below a non-exhaustive list of decisions that would generally be considered in scope of the transparency obligation: - facial recognition technology use in a retail store or stadium for watchlist matching - recruitment software used to sort through candidate profiles and make hiring decisions - differential and/or personalised pricing practices used by online retailers in the sale of significant goods - computer programs used to prioritise the provision of health or disability services to individuals - computer programs which determine eligibility for the provision of a government benefit or housing assistance - computer programs which determine eligibility for admission to an education or training program - computer programs used to approve or reject a loan or credit application - programmatic advertising relating to the decision to sell significant goods or services - computer programs involved in assessing an individual’s eligibility for insurance - reports created by artificial intelligence used to rank employee performance, determine promotions, or decide bonus/remuneration outcomes - computer programs involved in decisions relating to the provision of a scholarship, grant or other financial support - computer programs used to process immigration applications or enforce border controls at an airport. September 2026 Page 6 Appendix 2: APP 1.7-1.9 Transparency Obligation Examples Example 1 – Scope of obligation and meaning of computer program Wombat is an APP entity who provides health and aged care services. It uses a spreadsheet in a computer program called Echidna to collate personal information of existing and prospective clients to triage individuals and identify patients who require immediate support. A formula in the spreadsheet scores individuals’ need for medical or aged care services based on a range of attributes relating to the personal information they have provided to Wombat. This includes an individual’s date of birth, address, marital status, distance from family members, and a range of other health information (such as blood pressure, body mass index, time since last medical checkup, and whether the individual takes medication). The spreadsheet is setup to automatically rank individuals based on their score. Wombat employees use the Echidna spreadsheet to decide which individuals to contact and schedule medical appointments or provide aged care support based on the rankings determined by the spreadsheet. The use of the Echidna spreadsheet constitutes a decision within the scope of the transparency obligation because the formulas of the computer program determine which individuals are provided with health and aged care services and support. The services provided by Wombat are significant goods or services which could reasonably be expected to affect the rights or interests of an individual. Importantly, if the rankings in the spreadsheet were used by Wombat staff to work out which individuals not to contact to offer support, this would still be in scope of the transparency obligation on the basis that the computer program would be used to refuse or fail to make a decision or do a thing. Example 2 – Meaning of arranged for Yarra Springs is an APP entity who provides affordable housing to individuals living in Gippsland, Victoria. Yarra Springs only has a limited number of dwellings available to rent to individuals in the local area and decides to procure a new software- as-a-service (SaaS) program to assist in making decisions about which individuals to provide housing to next. Yarra Springs selects a SaaS product called Residence which uses AI and machine learning capabilities to process data on housing applicants in real time, undertake predictive analytics on what type of housing individuals might need, and assist with making housing decisions by matching applicants needs with new dwellings that become available to rent. While Yarra Springs has paid for a premium version of Residence which allows it to have greater control over decision making parameters, the software also has a cheaper version which purely generates recommendations for housing decision. Regardless of whether Yarra Springs has the ability to customise or configure the decision parameters provided to Residence, it still uses the third -party software to make decisions in a September 2026 Page 7 manner which significantly affects the rights or interests of an individual (i.e. access to housing). On this basis, Yarra Springs has ‘arranged for’ Residence to be used in a manner which falls within the scope of the transparency obligation and is required to provide information of the kinds of personal information used and decisions made in its privacy policy. Example 3 – Meaning of substantially and directly related Esperance is an APP entity which sells technology products and software solutions to enterprise customers. The company has a sales team of 58 people within Australia who are remunerated based on key performance indicators including revenue targets, units sold, conversion rates, customer feedback, compliance training requirements, regulatory compliance, peer feedback, office attendance and contribution to team objectives. Managerial staff at Esperance use a generative AI program called GPTea to assist with completing employee performance reviews, awarding year -end bonuses and determining salary adjustments at the end of each financial year. The computer program is used to generate profiles on each sales staff member based on all key performance indicators and provides a suggested bonus amount and salary adjustment in a report which managerial staff use to make remuneration decisions. Esperance’s human resources team always have oversight and managerial staff must receive director approval on all final remuneration decisions in writing. Regardless of whether a human intervenes to override the output of the computer program, the recommen dation provided by GPTea is relied upon to document the reasoning behind remuneration decisions. Given Esperance’s managerial staff use the output of GPTea as a key factor in making a remuneration decision, which has a direct connection to individual sales staff’s remuneration, the generative AI output is substantially and directly related to Esperance’s decision-making. However, if Esperance staff undertook additional processes such as interrogating outputs, conducting further review of individual key performance indicators behind the decision, setting GPTea decision parameters more narrowly , using additional evidence to come to a remuneration decision, and documenting the reasoning why it diverged from the recommendation of the computer program, this could indicate that the output was not substantially related to the making of an individual decision. Example 4 – Meaning of significantly affects rights and interests (a ccess to goods and differential/personalised pricing) Daintree is an e-commerce platform that allows subscribers to purchase essential goods to their home. When signing up to Daintree, users are required to disclose their address. Daintree adjusts pricing of products on its platform depending on the subscriber’s home address and the level of demand for specific goods within a postcode. Daintree does not September 2026 Page 8 charge delivery fees and price charged to the subscriber is the same as advertised to the subscriber. Priya, who lives in Limetree Place in Sydney’s Inner West, is a Daintree subscriber and purchases a basket of groceries from the platform each week including a range of significant goods (including baby formula, gluten free foods, pharmacy items and over -the-counter medications). Priya meets up with her friend, Sophie, a single mum who lives in Skyline City in Sydney CBD and who is also a Daintree subscriber. Priya expresses how she feels the price of her groceries seems to be always going up and is surprised to learn that Sophie has been paying $0.89 to $3.72 less per item due to Daintree’s use of differential pricing to change prices depending on a subscriber’s postcode. Priya and Sophie compare their purchases of baby formula and calculate that Priya has paid $132 more than Sophie over the past year. Priya and Sophie also look through personal advertising they have received on other essential goods and find the same car seat they were looking at for their babies advertised to them on Daintree with a $102 difference. The use of differential pricing by Daintree’s computer programs is significant regardless of whether the price difference is small or large. As this example demonstrates, even a small difference can compound over time to be a significant quantum which is more than trivial. Daintree’s use of computer programs to provide differential pricing also affect the rights or interests of Priya and Sophie because it impacts their access to essential goods such as food and medication. Under the transparency obligation, Daintree would be required to provide information about how it uses personal information to make differential pricing decisions in a way which impacts individuals’ rights or interests. Example 5 – Meaning of significantly affects rights and interests (recruitment software ) Manscon, an engineering firm, is hiring for graduate engineers. Manscon advertises open roles on recruitment site, Opaque Window, to attract the best candidates to the firm. Opaque Window algorithmically promotes this job posting to candidates it thinks ar e best suited. One metric it uses is the age and gender of account holders. Since most engineers at Manscon are young males, Opaque Window prioritises promoting the job to male engineering graduates aged 21 to 30. Rafqa (aged 45) is a female engineering graduate who has just finished her degree following her decision to change careers five years ago. She is now searching for her first engineering job on Opaque Window. She does not receive the Manscon job ad due to the targeting settings of the computer program, which limits her employment options. Nevertheless, Rafqa hears about the role from her close friend, Carlos , who saw the ad on Opaque Window, and applies for the role. Rafqa and Carlos receive an email from Manscon stating that the firm will be using an AI - powered screening tool from Opaque Window called Monologue to assist with the next stage September 2026 Page 9 of the graduate recruitment process. Rafqa and Carlos both sit a first -round digital, pre- recorded video interview where they record their answers to a blank -screen which is uploaded to Manscon’s recruitment portal. Monologue creates an AI-generated candidate profile on each individual who applies for a role based on inputs it receives from an individuals’ resume and an analysis of answers provided during a first-round video interview. The tool scores an applicant’s performance in the video interview based on a range of factors including biometric inputs such as the tone of their voiceprint, speech patterns, facial geometry, eye movement, and body language. Monologue produces a summary report and ranks applicants into a shortlist for Manscon’ s HR team to contact candidates for a final round interview. Manscon’s use of Opaque Window and its companion software Monologue are in scope of the transparency obligation on the basis that they significantly affect the rights or interests of individuals. In this regard, Opaque Window significantly affects the rights or interests of Rafqa because the computer program has the ability to limit her access to employment opportunities. Information about Manscon’s use of Monologue must also be included in its privacy policy on the basis that the software uses personal information in a w ay to make automated decisions about which candidates are selected for a final round interview. Example 6 – Meaning of commercial-in-confidence Banksia is a financial institution that provides individuals with access to loans or credit for the purpose of purchasing cars, trucks or machinery in regional Western Australia. Banksia uses a range of computer programs to detect fraud based on individuals’ personal information . The use of individuals’ personal information to power fraud detection tools is standard across the financial services industry to prevent access to individuals accounts and does not confer a competitive advantage on Banksia. Recently, Banksia has developed a new in-house computer program called ‘Noah’ to identify advanced patterns of fraudulent activity based on individuals’ personal information in the course of making decisions about approving or rejecting a loan or credit application. The ‘Noah’ computer program is unique to Banksia because it uses cutting -edge blockchain technology connected with explainable AI tools to analyse individuals’ personal information (including name, employment record information, credit information, biometric templates and location information). The program draws on this source data and assigns a specific weighting to each data point to flag unusual or suspicious activity in loan applications and generate credit scores on individuals with no credit history. This has been instrumental in allowing Banksia to make credit and loan decisions more efficiently and has introduced a new demographic of customers to the institution. Banksia has received significant offers from global competitors to purchase ‘Noah’ technology and is concerned that if it includes information about the computer program’s use of personal information to make credit and loan decisions, this could be reverse engineered by a competitor and reduce its competitive advantage. September 2026 Page 10 Information about how Noah weights data points to flag fraudulent activity can be classified as commercial-in-confidence and does not need to be included in Banksia’s privacy policy. However, the use of individuals’ personal information to power fraud detection tools is not commercial-in-confidence. On this basis, Banksia details the following information which is compliant with the transparency obligation in a new section within its privacy policy: Decisions made by computer programs Banskia operates a combination of different computer programs to detect and prevent fraud which require the use of the following kinds of personal information: - identity information (e.g. name) - geolocation data - biometric templates (e.g. face scans) - employment record information - credit information. This information is used to make the following kinds of decisions by Banksia: - grant or deny access to personal banking accounts - grant or deny the processing of a transaction - grant or deny the processing of a loan or credit application . Example 7 – Extent of disclosure (high volume of automated decisions) The National Insurance Agency (NIA) makes decisions about eligibility for individuals accessing government support payments and training services when they are unable to work. NIA has recently procured a new software platform called SwiftEmu to make eligib ility assessments quicker based on personal information provided by individuals. SwiftEmu processes this information to generate a report for NIA staff which summarises eligibility for different support payments and training programs, automatically calculates the total benefit that the individual is entitled to, and identifies upcoming training courses which the individual can be enrolled in. NIA human staff always have oversight and make the final decision about eligibility based on the output provided. While NIA staff interface and engage with through SwiftEmu, the single report ultimately generated by the platform combines a range of significant decision outputs made by different computer programs that operate in the background. Some of these decisions are made in a sequential order (e.g. eligibility for different payments occur in rapid succession) whilst others are integrated into the platform itself (e.g. identification of different training courses). September 2026 Page 11 In complying with APP 1.7-1.9 obligations, NIA demonstrates compliance with the transparency obligation by providing the following information in a new section of its privacy policy: Decisions made by computer programs NIA operates computer programs to support the delivery of government support payments. This requires the use of the following kinds of personal information: - name - address - date of birth - employee record information - health information. This information is used by NIA to support decision -making and administrative action, workflow automation, and service delivery to make the following kids of decisions: - determine eligibility for government support payments - approve or deny government support payments . NIA also operates computer programs to support the delivery of training services to individuals. This requires the use of the following kinds of personal information: - education history - location information - criminal record. This information is used by NIA to support decision -making and administrative action, workflow automation, and service delivery to make the following kids of decisions: - generate individual candidate profiles - determine eligibility for education and training programs - identity educational and training programs that may be suitable for individuals - enrol individuals in approved educational or training programs .