Your privacy training is from 2022. Your AI tools are from last week.
Most Australian organisations have a privacy training course. It covers the Privacy Act 1988, the thirteen Australian Privacy Principles, what counts as personal information, what to do in a breach. It is competent, dry, and it was almost certainly written before generative AI tools became the default way half your staff draft, summarise, and analyse anything that lands in their inbox.
That gap is now a problem. The 2024 amendments to the Privacy Act tightened the rules around automated decision-making and added explicit AI-related obligations. The Australian Voluntary AI Safety Standard added another layer. The OAIC has been increasingly active on AI-related guidance. And in the meantime, your staff are pasting things into AI tools that they would never email externally.
A custom AI privacy training course closes the gap between what your privacy framework says and what your staff are actually doing.
This post is for the privacy officer, compliance lead, or P&C director who knows the existing privacy training is not enough but is not sure what a fit-for-2026 version should actually look like.
Why generic privacy training falls short
The standard privacy training course covers the APPs at a high level. It explains the difference between personal and sensitive information. It walks through breach notification obligations. It is, broadly, the same course it has been for a decade.
What it does not cover:
- What “use” of personal information means when you paste a customer record into ChatGPT
- Whether AI-generated content based on personal data counts as a new disclosure
- How the cross-border disclosure principle (APP 8) applies to AI tools with offshore processing
- What “reasonable steps” to protect personal information looks like when the tool is in the cloud and you do not control the infrastructure
- How automated decision-making provisions affect AI-assisted hiring, performance management, or customer service
- What to do when an AI tool hallucinates personal information about a real person
These are not edge cases. They are the day-to-day reality of any organisation that has rolled out AI in the last eighteen months. And none of them are covered by the privacy course your staff completed two years ago.
What a good AI privacy training course should cover
A custom AI privacy course should do five things.
1. Map AI use to the APPs explicitly
The Australian Privacy Principles are the framework. Most staff have seen them before. The training should not re-teach them. It should map them to AI use cases.
APP 6 (use and disclosure): does pasting customer data into an AI tool count as use, disclosure, or both? APP 8 (cross-border): which AI tools route data offshore, and what does that mean for your obligations? APP 11 (security): what counts as a reasonable security step when the tool is a third-party SaaS product?
The training should walk through each relevant APP and show the AI-specific question it now raises. Concrete examples, not abstract framing.
2. Cover the 2024 amendments
The Privacy and Other Legislation Amendment Act 2024 changed the picture. New obligations around automated decision-making. New transparency requirements. New direct right of action for individuals.
Your staff do not need to memorise the legislation. They need to understand which of their AI-related decisions are now in scope and what that practically means for how they work. The training should translate the legislation into role-relevant behaviour.
3. Address the cross-border problem
Most AI tools your staff are using process data outside Australia. ChatGPT routes through the US. Claude processes in the US and EU. Gemini, the same. Copilot depends on the licence configuration.
APP 8 puts obligations on you when personal information goes offshore. The training should make this concrete: which tools are configured to keep data in Australia, which are not, and what your staff should and should not paste into each. This is also why AI privacy training pairs naturally with Cyber Security Awareness training: the two categories overlap significantly when AI tools are in scope.
This is the section where generic privacy training fails hardest. It cannot name tools. Yours can.
4. Cover the hallucinated personal information problem
This is the AI-specific risk that most training ignores. AI tools fabricate. Sometimes they fabricate personal information about real, identifiable people: names, employment histories, qualifications, criminal records.
When that fabricated information ends up in a document your organisation produces, you have a problem. Possibly a breach. Possibly defamation. Definitely a complaint.
Staff need to know that AI-generated information about identifiable people must be verified before it leaves the building. The training should make this rule concrete and stick it in a place learners will remember.
5. Connect to your breach response process
If something goes wrong, what happens? Who do staff tell? How fast? What do they preserve?
Your existing privacy training probably covers this for traditional breaches. The AI version needs to extend it: what counts as an AI-related privacy incident, how to recognise one, who to escalate to. This is the slide that prevents the small problem becoming a notifiable one.
How long should the course be?
Fifteen to twenty minutes for general staff. Thirty minutes if you have a more sophisticated audience: privacy champions, legal, senior P&C.
Most organisations need two versions: a short one for the whole workforce, and a longer one for the people who handle privacy decisions day-to-day. We typically build them as a pair, with the longer version reusing 60-70 percent of the shorter one’s content.
What it should cost
AI privacy training built properly is Tier 2 territory: $7,500 plus GST. The reason is not visual complexity. It is the legal accuracy required. Privacy training that gets the legislation wrong is worse than no training. The course needs proper legal review, scenarios that hold up to scrutiny, and assessment that proves comprehension rather than just clicks.
If you want a slimmer awareness-only version, Tier 1 at $5,500 is workable for the all-staff audience, with the legal review built in.
If you need the dual-version approach, short for all staff and longer for privacy champions, that sits at Tier 3 at $10,500 because you are effectively getting two related courses that share content.
When to commission it
Now, if you have rolled out AI tools. Before rollout, if you have not.
The Privacy Act amendments are in effect. The OAIC is increasingly active. Insurers are starting to ask. If your privacy training is silent on AI in 2026, you are out of step with the regulatory direction and the risk reality both.
What to brief your developer on
When you commission AI privacy training, bring:
- Your existing privacy training. So the new course extends rather than duplicates.
- Your AI tool inventory. What is licensed, what is permitted on personal accounts, what is blocked.
- Your data classification scheme. What counts as personal, sensitive, confidential in your context.
- Your breach response process. So the course can point to the right escalation path.
- Legal review capacity. Either internal or external. The course will need it.
The bigger picture
AI privacy training sits at the intersection of two compliance pillars that have historically been treated separately: Privacy & Data Protection training and the emerging AI training category. Most organisations are realising that they cannot keep these separate any more. The AI tools have collapsed the distinction.
This is the third piece of the AI capability stack. AI acceptable use training covers the everyday behaviour. AI training for managers covers the leadership layer. This one covers the regulatory layer.
When you are ready, get in touch. Three weeks from kickoff to launch. Fixed price. No drama.