Australian organisations are adopting AI faster than they are updating their privacy practices. The assumption that the Privacy Act somehow does not reach new technology is both common and wrong. There is no AI carve-out. The moment an AI system touches personal information, the Australian Privacy Principles apply in full, and the consequences of getting it wrong are real.
The Privacy Act applies to AI, fully
The Privacy Act 1988 (Cth) and its thirteen Australian Privacy Principles (APPs) regulate how APP entities handle personal information, which is information about an identified individual or one who is reasonably identifiable. AI does not change that definition or those obligations. If you collect, use, store or disclose personal information through an AI tool, every relevant APP still binds you.
The APPs apply to APP entities: most Australian Government agencies and private-sector organisations with annual turnover above three million dollars, along with certain others such as health service providers and businesses that trade in personal information. If your organisation is regulated under the Act for its existing systems, nothing about adding AI removes that coverage. If anything, AI tends to expand the volume and sensitivity of personal information in play.
The APPs most affected by AI
While all the principles can be engaged, several deserve particular attention when AI enters the picture.
APP 1: Open and transparent management
You must manage personal information in an open and transparent way, including by maintaining a clear, up-to-date privacy policy. If AI now influences how you collect or use personal information, your privacy policy needs to reflect that honestly rather than describe a pre-AI world.
APP 3 and APP 5: Collection and notice
You may only collect personal information that is reasonably necessary for your functions, and you must notify individuals about the collection. Training or fine-tuning a model on customer data, or routing intake through an AI agent, are collection and use activities that your notices need to cover. Repurposing data collected for one reason to feed an unrelated AI use can breach the primary-purpose rule.
APP 6: Use and disclosure
Personal information collected for one purpose generally cannot be used or disclosed for another without consent or an exception. Feeding existing records into a new AI tool is a fresh use that must be justified against the original purpose. This is one of the most common ways well-meaning AI projects drift into non-compliance.
APP 8: Cross-border disclosure
Many AI providers process data overseas. Under APP 8, before disclosing personal information to an overseas recipient you must take reasonable steps to ensure they handle it consistently with the APPs, and in most cases you remain accountable for their conduct. You cannot meet this obligation if you do not even know which country your data is processed in, which is the reality for a great deal of shadow AI use.
APP 10 and APP 11: Quality and security
You must keep personal information accurate and protect it from misuse, loss and unauthorised access. AI raises distinct quality concerns, because models can generate plausible but wrong information about real people, and distinct security concerns, because data flows to new third parties and integrations.
Sensitive information and automated decisions
Sensitive information, including health, biometric, racial, religious and similar data, attracts stronger protections and generally requires consent to collect. AI tools that infer sensitive attributes, or biometric systems such as facial recognition, sit squarely in this higher-risk territory and demand careful assessment.
Australian privacy reform is also moving toward greater transparency around automated decision-making. Organisations using AI to make or substantially influence decisions that affect individuals should anticipate clearer obligations to explain those decisions, and should build that transparency in now rather than retrofitting it later. The Office of the Australian Information Commissioner (OAIC) has published guidance on using commercially available AI products and on training generative AI models, and following that guidance is a sensible baseline.
Common privacy traps in AI projects
Most privacy failures in AI are not exotic. They are predictable missteps that a little foresight would have caught. The recurring ones include:
- Repurposing existing data. A dataset collected years ago for billing is fed into a new AI model for marketing or analytics. Unless that secondary use falls within the original purpose or an exception, it likely breaches APP 6.
- Assuming de-identification is permanent. Data that looks anonymous can often be re-identified, especially when combined with other datasets. Treating weakly de-identified data as outside the Act is a common and costly error.
- Ignoring inferred data. AI does not just store what you give it; it infers new attributes. A model inferring health, financial or behavioural traits is generating fresh personal, sometimes sensitive, information about individuals.
- Overlooking the offshore question. Teams adopt a tool without ever establishing where data is processed, which makes APP 8 compliance impossible to demonstrate.
- No human in the loop. Fully automated decisions affecting people, with no avenue for review or explanation, invite both privacy and fairness complaints.
Avoiding these traps is far cheaper than remediating them after a complaint to the OAIC or a notifiable data breach.
Practical steps to stay compliant
Compliance is achievable with deliberate process. The following steps cover most organisations.
- Map your AI data flows. Document what personal information enters which AI tools, where it is processed, who can access it, and how long it is retained. You cannot assess what you have not mapped.
- Run a privacy impact assessment. For any AI initiative involving personal information, complete a privacy impact assessment before go-live. The OAIC strongly recommends this for higher-privacy-risk projects, and for agencies it is effectively expected practice.
- Lock down public tools. Prohibit personal information in public, consumer-grade AI services and provide a governed enterprise alternative with contractual protections, no training on your data, and known data locations.
- Fix your notices and policies. Update your privacy policy and collection notices to cover AI use, overseas processing and any automated decision-making, in plain language.
- Assess your vendors. Treat AI providers as you would any data processor. Scrutinise their terms, security, data location and retention. This is core cyber due diligence, not a formality.
- Govern it on an ongoing basis. Privacy is not a launch-day checkbox. Build AI into your data breach response plan and review arrangements as tools and regulation evolve.
What a privacy breach can cost
The stakes have risen sharply. The Privacy Act now carries substantially increased penalties for serious or repeated interferences with privacy, and the Notifiable Data Breaches scheme requires you to notify the OAIC and affected individuals when an eligible breach is likely to result in serious harm. An AI-driven leak of personal information can trigger both. Beyond regulatory penalties, the reputational damage of being seen to mishandle citizens’ or customers’ data through an unassessed AI tool can far exceed any fine, particularly for government agencies and financial-services firms whose entire relationship with the public rests on trust. The economics strongly favour getting the controls right before deployment rather than after an incident. It is also worth remembering that obligations follow the data, not the tool: outsourcing processing to an AI vendor does not outsource your accountability under the Act. If a provider mishandles personal information you disclosed to them, your organisation is typically still the one answering to the regulator and to the individuals affected.
Privacy is part of a wider AI governance picture
Privacy compliance does not exist in a vacuum. It works best as one strand of a coherent approach to AI governance for Australian organisations, sitting alongside security, accuracy and accountability controls. Government agencies and regulated entities such as those under APRA CPS 234 face additional layers, but the privacy foundations described here apply to everyone.
At CISO Advisory Australia we help organisations adopt AI without tripping over their privacy obligations, from data-flow mapping and privacy impact assessments to vendor review and policy uplift. If you want a clear read on where your AI use stands against the Privacy Act, our AI security consulting team can help, or call us on 07 2112 8502. The Act has not changed for AI. Your practices need to.
Frequently asked questions
Does the Privacy Act apply to AI?
The Privacy Act 1988 does not single out AI, but it applies fully whenever AI systems collect, use, store or disclose personal information. There is no AI exemption. If your AI tool touches personal information, the Australian Privacy Principles apply exactly as they would to any other system handling that data.
Who must comply with the Australian Privacy Principles?
The APPs apply to APP entities, which include most Australian Government agencies and private-sector organisations with an annual turnover above three million dollars, plus some smaller organisations such as health service providers. Many smaller businesses are also caught through specific activities like handling health information or trading in personal information.
Can we put personal information into ChatGPT or similar tools?
Only with great care. Doing so is a use and often an overseas disclosure of personal information that must comply with the APPs. Public tools rarely offer the contractual protections you need, so most organisations should prohibit personal information in public AI tools and provide a governed alternative instead.
What is the risk with overseas AI providers?
Many AI services process data offshore. Under APP 8, disclosing personal information to an overseas recipient generally makes you accountable for their handling of it unless an exception applies. You need to know where data goes, what protections exist, and whether your privacy notices and consents cover it.
Do we need to update our privacy policy for AI?
Almost certainly. APP 1 and APP 5 require transparency about how you handle personal information. If AI now plays a role in collection, use or decision-making, your privacy policy and collection notices should say so in clear terms, including any overseas processing and automated decision-making.
Need this handled for your organisation?
Confidential and no obligation. We respond the same business day — on-site same day / next business day, or remote, Australia-wide. Prefer to talk now? Call us 24/7 on 07 2112 8502.