Using AI With Customer Data: A Privacy Act Checklist for Australian Businesses

Calls, emails, enquiry forms and CRM records are full of personal information. Once an AI system reads, summarises or writes it, the Privacy Act 1988 comes with it. This checklist is for Australian businesses adopting chatbots, voice agents or AI employees connected to their inbox and CRM.

Does the Privacy Act apply to your business?

Start here, because many small businesses are exempt. According to the OAIC's small business guidance, most businesses with an annual turnover of $3 million or less are not covered. Turnover includes all income from all sources.

The Act covers some businesses regardless of turnover, including:

• health service providers;
• businesses that trade in personal information;
• contractors providing services under a Commonwealth contract;
• credit reporting bodies and residential tenancy database operators;
• reporting entities under the AML/CTF Act;
• businesses related to a covered business, and businesses that opt in.

Part IIIA of the Privacy Act, which regulates consumer credit information such as credit reports, can also apply. The OAIC publishes a checklist to help you decide. Even if you are exempt, the checklist below is still good practice for handling customer data.

What does the OAIC say about using AI products?

The OAIC's guidance on privacy and the use of commercially available AI products (published 21 October 2024, updated 17 January 2025) makes five headline points:

• Privacy obligations apply to personal information input into an AI system and to output data that contains personal information.
• Do due diligence before adopting a product: whether it has been tested for your use, how human oversight fits, the privacy and security risks, and who will have access to the information.
• Update privacy policies and notices, and clearly identify public-facing AI tools, such as chatbots, as AI.
• Generating or inferring personal information with AI is a collection under APP 3, and inferred or incorrect information about an identifiable person, including hallucinations, is personal information.
• Under APP 6, personal information can only be used for the purpose it was collected for, unless the person consents or would reasonably expect the related secondary use.

As best practice, the OAIC recommends not entering personal information, particularly sensitive information, into publicly available generative AI tools.

Which Australian Privacy Principles matter most for AI?

There are 13 Australian Privacy Principles. The OAIC's AI guidance focuses on these:

APP	What it covers	What it means for AI
APP 1	Open and transparent management	Privacy policy explains AI use; automated-decision details from 10 December 2026
APP 3	Collection	AI-generated or inferred details about a person count as collection
APP 5	Notification	Tell people at collection, for example that a call is recorded and transcribed
APP 6	Use and disclosure	Feeding data into AI for a new purpose may be a secondary use
APP 8	Cross-border disclosure	Know which countries your AI provider and model providers process data in
APP 10	Quality	Take reasonable steps to keep AI output about people accurate
APP 11	Security	Protect information from misuse, loss and unauthorised access

What changes on 10 December 2026?

The Privacy and Other Legislation Amendment Act 2024 added a transparency rule for automated decisions to APP 1. The OAIC confirmed on 30 September 2026 that from 10 December 2026, APP entities must include information in their privacy policy where they have arranged for a computer program to make, or do something substantially and directly related to making, a decision that could reasonably be expected to significantly affect an individual's rights or interests, using that person's personal information.

The policy must describe the kinds of personal information used and the kinds of decisions involved. The OAIC has published a fact sheet, a flowchart and updated APP 1 guidelines. An AI that only drafts a reply for a person to approve is different from one that decides who gets a service, so map each AI workflow against the test. The same 2024 reforms also introduced a statutory tort for serious invasions of privacy and expanded the OAIC's enforcement powers.

What if an AI tool causes a data breach?

Under the Notifiable Data Breaches scheme, covered organisations must notify affected individuals and the OAIC when a breach is likely to result in serious harm. The OAIC's examples include personal information mistakenly given to the wrong person, which is exactly the risk when an AI sends an email or summary to the wrong contact. If you suspect an eligible breach, you must take all reasonable steps to complete an assessment within 30 calendar days, and the OAIC expects it to be done much faster where possible. Logs of what the AI read and sent make that assessment far quicker; our AI security and supervision FAQ covers what to log.

The Privacy Act checklist for using AI with customer data

1. Confirm coverage. Work out whether the Privacy Act and Part IIIA apply, using the OAIC's small business checklist.
2. Map the data. List which systems the AI connects to and which personal information it will read, generate and store, including recordings and transcripts.
3. Check purpose. For each workflow, confirm the use matches why you collected the information, or that people would reasonably expect it (APP 6).
4. Do provider due diligence. Get written answers on where data is processed, model training, retention, deletion, sub-processors and staff access.
5. Check overseas disclosure. Identify any offshore processing and how APP 8 is met.
6. Keep personal information out of public AI tools. Set a staff rule and back it with training, as the OAIC recommends.
7. Update notices and your privacy policy. Explain AI use, identify AI chatbots and voice agents as AI, and announce call recording.
8. Assess automated decisions. Before 10 December 2026, check whether any AI workflow meets the APP 1.7 test and update your policy if so.
9. Build in accuracy checks. Use human review for anything sent to customers or written to a person's record (APP 10).
10. Limit access and secure it. Give the AI least-privilege access and turn on MFA for connected accounts (APP 11).
11. Plan for breaches. Add AI scenarios to your data breach response plan, including how to pause the AI and review its logs.
12. Consider a privacy impact assessment. The OAIC recommends one as part of privacy by design for AI products.
13. Review regularly. The OAIC says due diligence should not be "set and forget".

Industry rules can add to this: see our guides for law firms using AI summaries and mortgage and finance brokers, and our guide to AI phone call compliance for recording and telemarketing. When Wharq scopes an AI employee, the audit covers the data it will touch and confirms hosting specifics in writing; the roles are on our services page.

AI and the Privacy Act: frequently asked questions

Can staff paste customer details into ChatGPT or other public AI tools?

The OAIC recommends, as best practice, that organisations do not enter personal information, particularly sensitive information, into publicly available generative AI tools. Set a clear staff rule and give people an approved alternative.

Is an AI call transcript personal information?

Yes, if it is about an identified or reasonably identifiable person. Recordings, transcripts, summaries and any details the AI infers about the caller should be handled under the APPs if your business is covered.

We turn over less than $3 million. Can we ignore the Privacy Act?

Not automatically. Health service providers, businesses that trade in personal information, Commonwealth contractors, AML/CTF reporting entities and some others are covered regardless of turnover, and Part IIIA credit rules can apply. Use the OAIC checklist to confirm.

Does an AI that drafts emails count as automated decision-making?

Usually the concern is decisions that could significantly affect someone's rights or interests, such as eligibility for a service. A draft that a person reviews and sends is different, but map each workflow against the APP 1.7 test in the OAIC's fact sheet.

What should we ask an AI provider about privacy?

Ask where data is processed, whether inputs or outputs are used for training, how long data is kept and how it is deleted, which sub-processors receive it, and who can access it.

Sources

1. OAIC, "Small business" (accessed October 2026)
2. OAIC, "Guidance on privacy and the use of commercially available AI products" (published 21 October 2024, updated 17 January 2025)
3. OAIC, "Australian Privacy Principles" (accessed October 2026)
4. Federal Register of Legislation, "Privacy and Other Legislation Amendment Act 2024" (No. 128, 2024; accessed October 2026)
5. OAIC, "New resources on transparency for use of AI and automated decision-making" (published 30 September 2026)
6. OAIC, "Passing of bill a significant step for Australia's privacy law" (published 29 November 2024)
7. OAIC, "About the Notifiable Data Breaches scheme" (accessed October 2026)
8. OAIC, "Part 4: Notifiable Data Breach (NDB) Scheme" (accessed October 2026)