---
title: "Privacy, Data Protection & Cyber"
url: "https://ailawframework.com/practice/privacy-data-protection-and-cyber"
type: "practice_area"
site: "AI Law Framework"
id: "A25"
task_count: 32
last_reviewed: "2026-10-10"
---

# Privacy, Data Protection & Cyber

> Establish the actual information, processing activity and incident facts before choosing a response. Verify applicable duties and time limits, and keep access, training use, retention and disclosure conditions distinct.

Canonical page: https://ailawframework.com/practice/privacy-data-protection-and-cyber  
Last reviewed: 2026-10-10

Site index for agents: https://ailawframework.com/llms.txt

Also covers: Privacy law; data breach response; cyber incident law; information security law; surveillance law; data governance advisory

## Key facts

- Tasks: 32
- Green: 3, Yellow: 24, Red: 5
- Levels: Green is AI drafts and the lawyer checks; Yellow is lawyer and AI draft together; Red is the lawyer writes and AI may only check, clarify or collect data.

## Principles

- **Define the matter**: Confirm the client’s instructions, jurisdiction, intended output and the information available before dividing the work.
- **Verify the authorities**: Open each relied-on authority independently. Confirm its existence, current status, proposition and any quotation or pinpoint reference.
- **Reconcile the figures**: Trace every relied-on figure to its source, confirm dates and assumptions, and independently recalculate totals or derived amounts.

## Checklist

- Open each relied-on authority independently. Confirm its existence, current status, proposition and any quotation or pinpoint reference.
- Trace every relied-on figure to its source, confirm dates and assumptions, and independently recalculate totals or derived amounts.
- Calculate operative dates independently from the current rule and actual triggering event. Check exceptions and record the verified deadline.
- Confirm that the actual tool, feature and processing terms permit use of this information, including access, retention, training and disclosure conditions.
- Verify the applicable jurisdiction and the version of each relevant law, rule or instrument in force at the material time.
- Resolve exceptions and record the responsible lawyer’s approval of the exact output and intended use.

## Restrictions

- Confirm firm, client and applicable court or professional restrictions before using any AI feature.

## Jurisdiction note

Apply this guidance to the actual matter and jurisdiction. Confirm current requirements before relying on an example.

## Basis

Practice guidance supports planning; the responsible lawyer confirms the scope, evidence and review requirements for the matter.

Also covers: Privacy law; data breach response; cyber incident law; information security law; surveillance law; data governance advisory

This practice area sits at the intersection of the Privacy Act 1988 (Cth) and its 2024 amendments, the Cyber Security Act 2024 (Cth), the Security of Critical Infrastructure Act 2018 (Cth), the Australian Privacy Principles (APPs), and the Notifiable Data Breaches (NDB) scheme. AI risk is exceptionally high here for two compounding reasons: (1) the subject-matter itself is complex, jurisdiction-layered and rapidly changing, making hallucinated authority directly exploitable in compliance advice; and (2) the act of using AI tools to handle client data may itself constitute the breach the firm is advising others to prevent — creating an additional confidentiality and governance risk.

## Limits on AI assistance

Red lines — never in this area

## Limits on AI assistance

AI must not independently determine whether a data breach meets the NDB 'likely to result in serious harm' threshold — this is a factual-legal judgment with direct notification obligations and penalty exposure up to the greater of $50m, 3x benefit, or 30% of adjusted turnover; a lawyer must apply it to the actual facts. AI must not draft or finalise any submission, response, or communication that will be filed with the OAIC, a court, or any regulator — these are Red-tier outputs regardless of how the task is framed; the entire text must be lawyer-authored. AI must not structure or sign off on any privilege claim over breach investigation materials — Singtel Optus Pty Ltd v Robertson (Full Federal Court, May 2024) and McClure v Medibank Private [2025] FCA 167 confirm that privilege is destroyed when reports serve non-dominant legal purposes; a lawyer must structure every retainer, every report instruction and every communication channel before the investigation begins. AI must not advise on whether the statutory tort of serious invasion of privacy (commenced 10 June 2025) applies to specific conduct without full lawyer re-derivation of all elements — the intentional/reckless threshold, proportionality balancing, limitation period (including minor plaintiff and court-extension provisions), and absence of any case law must all be lawyer-assessed. AI must not be used to process, analyse or prompt with actual client personal information (including breach notification data, health records, or dispute materials containing identifying information) in any public or external-server AI tool — this is itself an APP 11 breach, a potential NDB event, and, for health information, a potential state health privacy Act breach. AI must not cite OAIC determinations, Federal Court privacy decisions or regulatory guidance without independent verification on oaic.gov.au and austlii.edu.au — AI hallucination of OAIC determinations is a direct professional conduct risk; more than 20 Australian lawyers have faced regulatory proceedings since 2024 for submitting fake AI-generated citations. AI must not determine whether the Cyber Security Act 2024 ransomware payment reporting obligation applies to a specific entity — the 72-hour obligation turns on turnover ($3m threshold) and third-party payment awareness, both of which require confirmation on the actual facts; a missed mandatory report carries a civil penalty of $99,000 (body corporate). AI must not determine whether an entity holds a 'critical infrastructure asset' under the SOCI Act or what reporting timeframe applies — the 12-hour versus 72-hour distinction has significant civil penalty exposure and the 22 asset-class definitions require lawyer verification against technical sector criteria. AI must not produce a final opinion letter or compliance sign-off on APP compliance status — the regulatory risk of an incorrect clearance letter (particularly in light of the $5.8m Australian Clinical Labs penalty and OAIC's active enforcement posture) is direct client harm and professional indemnity exposure. AI must not draft communications to affected individuals in a data breach without a lawyer confirming the communication is structured to preserve privilege over the investigation and does not inadvertently admit facts that elevate the OAIC's enforcement options.

- AI must not independently determine whether a data breach meets the NDB 'likely to result in serious harm' threshold — this is a factual-legal judgment with direct notification obligations and penalty exposure up to the greater of $50m, 3x benefit, or 30% of adjusted turnover; a lawyer must apply it to the actual facts.
- AI must not draft or finalise any submission, response, or communication that will be filed with the OAIC, a court, or any regulator — these are Red-tier outputs regardless of how the task is framed; the entire text must be lawyer-authored.
- AI must not structure or sign off on any privilege claim over breach investigation materials — Singtel Optus Pty Ltd v Robertson (Full Federal Court, May 2024) and McClure v Medibank Private [2025] FCA 167 confirm that privilege is destroyed when reports serve non-dominant legal purposes; a lawyer must structure every retainer, every report instruction and every communication channel before the investigation begins.
- AI must not advise on whether the statutory tort of serious invasion of privacy (commenced 10 June 2025) applies to specific conduct without full lawyer re-derivation of all elements — the intentional/reckless threshold, proportionality balancing, limitation period (including minor plaintiff and court-extension provisions), and absence of any case law must all be lawyer-assessed.
- AI must not be used to process, analyse or prompt with actual client personal information (including breach notification data, health records, or dispute materials containing identifying information) in any public or external-server AI tool — this is itself an APP 11 breach, a potential NDB event, and, for health information, a potential state health privacy Act breach.
- AI must not cite OAIC determinations, Federal Court privacy decisions or regulatory guidance without independent verification on oaic.gov.au and austlii.edu.au — AI hallucination of OAIC determinations is a direct professional conduct risk; more than 20 Australian lawyers have faced regulatory proceedings since 2024 for submitting fake AI-generated citations.
- AI must not determine whether the Cyber Security Act 2024 ransomware payment reporting obligation applies to a specific entity — the 72-hour obligation turns on turnover ($3m threshold) and third-party payment awareness, both of which require confirmation on the actual facts; a missed mandatory report carries a civil penalty of $99,000 (body corporate).
- AI must not determine whether an entity holds a 'critical infrastructure asset' under the SOCI Act or what reporting timeframe applies — the 12-hour versus 72-hour distinction has significant civil penalty exposure and the 22 asset-class definitions require lawyer verification against technical sector criteria.
- AI must not produce a final opinion letter or compliance sign-off on APP compliance status — the regulatory risk of an incorrect clearance letter (particularly in light of the $5.8m Australian Clinical Labs penalty and OAIC's active enforcement posture) is direct client harm and professional indemnity exposure.
- AI must not draft communications to affected individuals in a data breach without a lawyer confirming the communication is structured to preserve privilege over the investigation and does not inadvertently admit facts that elevate the OAIC's enforcement options.

## Jurisdiction-specific considerations

Australian traps

## Jurisdiction-specific considerations

The Privacy and Other Legislation Amendment Act 2024 (Cth) received Royal Assent on 10 December 2024. First-tranche obligations commenced immediately: enhanced penalties up to $50m (greater of $50m / 3x benefit / 30% adjusted turnover), new tiered civil penalty structure ($3.3m body corporate for non-serious; $50m+ for serious or repeated), new direct marketing opt-out, infringement notice powers, enhanced OAIC enforcement powers. Tranche 2 — removal of the $3m small business exemption broadly, further individual rights — is NOT yet enacted as at June 2026. AI may conflate enacted and proposed provisions. Always verify on legislation.gov.au. AML/CTF Tranche 2 creates a separate Privacy Act on-ramp from 1 July 2026: legal professionals, accountants, real estate agents, conveyancers, and dealers in precious metals/stones who become AML/CTF reporting entities under the Anti-Money Laundering and Counter-Terrorism Financing Amendment Act 2024 must comply with the Privacy Act for AML/CTF-related personal information regardless of their turnover. The firm itself may be caught. OAIC has published updated AML/CTF privacy guidance — check oaic.gov.au for current version. The statutory tort for serious invasions of privacy commenced 10 June 2025 (Schedule 2, Privacy and Other Legislation Amendment Act 2024). It applies only to conduct after that date. Only individuals (not corporations) may sue; defendants can be individuals or corporations. Standard limitation period: the earlier of 12 months from plaintiff's awareness and 3 years from the invasion. Court may extend to an outer limit of 6 years. Minors: proceedings must commence before the plaintiff turns 21. Aggravated damages are not available; exemplary/punitive damages are available in exceptional circumstances. Damages cap: the greater of $478,550 or the maximum defamation non-economic loss award; economic loss is uncapped. No cases have yet been decided — all elements remain judicially untested. First civil penalty under the Privacy Act: Australian Information Commissioner v Australian Clinical Labs Limited (No 2) [2025] FCA 1224 — $5.8 million penalty ordered in October 2025. This is the current judicial anchor for penalty submissions and breach quantum advice. AI may not know this decision or may hallucinate its terms — verify on austlii.edu.au. Privilege over breach investigation reports is actively litigated and difficult to establish. Singtel Optus Pty Ltd v Robertson (Full Federal Court, May 2024): Deloitte forensic report not privileged because it was not prepared for the dominant legal purpose — it served board, PR and non-legal purposes. McClure v Medibank Private [2025] FCA 167: mixed results — factual investigation reports not privileged (APRA and forensic expert engaged directly without dominant lawyer involvement); communications to cybersecurity experts were privileged. The retainer structure, instruction pathway and purpose must be established before the investigation begins — not after. Three parallel mandatory reporting streams may apply simultaneously following a cyber incident: (1) NDB scheme — 30-day assessment window from awareness of a suspected breach, then notify OAIC and affected individuals if the threshold is met; (2) Cyber Security Act 2024 s 30 — entities with turnover over $3m must report ransomware or cyber extortion payments to ASD via cyber.gov.au within 72 hours of making or becoming aware of the payment; civil penalty $99,000 (body corporate); (3) SOCI Act — critical infrastructure asset holders must report critical cyber security incidents to ACSC within 12 hours and other reportable incidents within 72 hours. AI routinely omits the SOCI Act stream and mis-characterises the ransomware obligation as criminal. Automated decision-making (ADM) transparency obligation (APPs 1.7-1.9) does NOT commence until 10 December 2026. OAIC consultation on guidance closed 15 June 2026 — check whether final guidance has been published. Do not advise current compliance is required; do advise that AI systems being designed and deployed now should be built to comply. Anti-discrimination law obligations apply independently to algorithmic decisions right now. State-based health privacy legislation operates alongside the Commonwealth Privacy Act and is not displaced by it: Health Records Act 2001 (Vic), Health Records and Information Privacy Act 2002 (NSW), Health Records (Privacy and Access) Act 1997 (ACT). For health-sector clients these impose additional obligations and AI may apply only the Commonwealth regime. State surveillance device legislation applies to employee monitoring independently of the Privacy Act, with materially different notice and consent requirements across jurisdictions: Surveillance Devices Act 1999 (Vic), Workplace Surveillance Act 2005 (NSW), Surveillance Devices Act 2016 (SA), Surveillance Devices Act 1999 (WA). AI frequently collapses these into a single generic rule. The Cth Surveillance Devices Act 2004 also applies to federal law enforcement contexts. Sector-specific overlays AI commonly misses or conflates: APRA CPS 234 (information security — all APRA-regulated entities), APRA CPS 230 (operational risk including third-party/supply chain — from 1 July 2025; pre-existing service provider contracts transition by 1 July 2026), My Health Records Act 2012, Security of Critical Infrastructure Act 2018 (SOCI Act — 22 asset classes, 12/72-hour reporting), and the Telecommunications (Interception and Access) Act 1979 for telco-related data requests. OAIC 2025-26 regulatory action priorities explicitly name: advertising technology (pixel tracking), practices that erode privacy rights in the application of AI, excessive collection and retention, rental/property/credit reporting/data-broker sectors, and biometric/facial recognition technologies. Clients in these sectors face elevated and active enforcement risk. The OAIC's posture has shifted from guidance-first to enforcement-first following the 2024 Act's enhanced powers.

- The Privacy and Other Legislation Amendment Act 2024 (Cth) received Royal Assent on 10 December 2024. First-tranche obligations commenced immediately: enhanced penalties up to $50m (greater of $50m / 3x benefit / 30% adjusted turnover), new tiered civil penalty structure ($3.3m body corporate for non-serious; $50m+ for serious or repeated), new direct marketing opt-out, infringement notice powers, enhanced OAIC enforcement powers. Tranche 2 — removal of the $3m small business exemption broadly, further individual rights — is NOT yet enacted as at June 2026. AI may conflate enacted and proposed provisions. Always verify on legislation.gov.au.
- AML/CTF Tranche 2 creates a separate Privacy Act on-ramp from 1 July 2026: legal professionals, accountants, real estate agents, conveyancers, and dealers in precious metals/stones who become AML/CTF reporting entities under the Anti-Money Laundering and Counter-Terrorism Financing Amendment Act 2024 must comply with the Privacy Act for AML/CTF-related personal information regardless of their turnover. The firm itself may be caught. OAIC has published updated AML/CTF privacy guidance — check oaic.gov.au for current version.
- The statutory tort for serious invasions of privacy commenced 10 June 2025 (Schedule 2, Privacy and Other Legislation Amendment Act 2024). It applies only to conduct after that date. Only individuals (not corporations) may sue; defendants can be individuals or corporations. Standard limitation period: the earlier of 12 months from plaintiff's awareness and 3 years from the invasion. Court may extend to an outer limit of 6 years. Minors: proceedings must commence before the plaintiff turns 21. Aggravated damages are not available; exemplary/punitive damages are available in exceptional circumstances. Damages cap: the greater of $478,550 or the maximum defamation non-economic loss award; economic loss is uncapped. No cases have yet been decided — all elements remain judicially untested.
- First civil penalty under the Privacy Act: Australian Information Commissioner v Australian Clinical Labs Limited (No 2) [2025] FCA 1224 — $5.8 million penalty ordered in October 2025. This is the current judicial anchor for penalty submissions and breach quantum advice. AI may not know this decision or may hallucinate its terms — verify on austlii.edu.au.
- Privilege over breach investigation reports is actively litigated and difficult to establish. Singtel Optus Pty Ltd v Robertson (Full Federal Court, May 2024): Deloitte forensic report not privileged because it was not prepared for the dominant legal purpose — it served board, PR and non-legal purposes. McClure v Medibank Private [2025] FCA 167: mixed results — factual investigation reports not privileged (APRA and forensic expert engaged directly without dominant lawyer involvement); communications to cybersecurity experts were privileged. The retainer structure, instruction pathway and purpose must be established before the investigation begins — not after.
- Three parallel mandatory reporting streams may apply simultaneously following a cyber incident: (1) NDB scheme — 30-day assessment window from awareness of a suspected breach, then notify OAIC and affected individuals if the threshold is met; (2) Cyber Security Act 2024 s 30 — entities with turnover over $3m must report ransomware or cyber extortion payments to ASD via cyber.gov.au within 72 hours of making or becoming aware of the payment; civil penalty $99,000 (body corporate); (3) SOCI Act — critical infrastructure asset holders must report critical cyber security incidents to ACSC within 12 hours and other reportable incidents within 72 hours. AI routinely omits the SOCI Act stream and mis-characterises the ransomware obligation as criminal.
- Automated decision-making (ADM) transparency obligation (APPs 1.7-1.9) does NOT commence until 10 December 2026. OAIC consultation on guidance closed 15 June 2026 — check whether final guidance has been published. Do not advise current compliance is required; do advise that AI systems being designed and deployed now should be built to comply. Anti-discrimination law obligations apply independently to algorithmic decisions right now.
- State-based health privacy legislation operates alongside the Commonwealth Privacy Act and is not displaced by it: Health Records Act 2001 (Vic), Health Records and Information Privacy Act 2002 (NSW), Health Records (Privacy and Access) Act 1997 (ACT). For health-sector clients these impose additional obligations and AI may apply only the Commonwealth regime.
- State surveillance device legislation applies to employee monitoring independently of the Privacy Act, with materially different notice and consent requirements across jurisdictions: Surveillance Devices Act 1999 (Vic), Workplace Surveillance Act 2005 (NSW), Surveillance Devices Act 2016 (SA), Surveillance Devices Act 1999 (WA). AI frequently collapses these into a single generic rule. The Cth Surveillance Devices Act 2004 also applies to federal law enforcement contexts.
- Sector-specific overlays AI commonly misses or conflates: APRA CPS 234 (information security — all APRA-regulated entities), APRA CPS 230 (operational risk including third-party/supply chain — from 1 July 2025; pre-existing service provider contracts transition by 1 July 2026), My Health Records Act 2012, Security of Critical Infrastructure Act 2018 (SOCI Act — 22 asset classes, 12/72-hour reporting), and the Telecommunications (Interception and Access) Act 1979 for telco-related data requests.
- OAIC 2025-26 regulatory action priorities explicitly name: advertising technology (pixel tracking), practices that erode privacy rights in the application of AI, excessive collection and retention, rental/property/credit reporting/data-broker sectors, and biometric/facial recognition technologies. Clients in these sectors face elevated and active enforcement risk. The OAIC's posture has shifted from guidance-first to enforcement-first following the 2024 Act's enhanced powers.

## Risks to address

How it goes wrong

## Risks to address

Hallucinated OAIC determinations: AI invents plausible-sounding case names and holdings that do not exist. In a regulatory response to the OAIC this causes immediate professional conduct exposure. Since 2024 more than 20 Australian court incidents have been recorded; the VLSB+C varied a solicitor's practising certificate in August 2025 following AI-citation misconduct in family law proceedings. Treating proposed law as enacted: AI conflates Tranche 2 Privacy Act reforms (broad small business exemption removal, right to erasure, fair and reasonable test) with the enacted 2024 first-tranche changes. Separately, AI may miss the AML/CTF Tranche 2 on-ramp from 1 July 2026 that catches lawyers and other professional services providers regardless of turnover. Missing the ransomware reporting penalty category: AI may describe the Cyber Security Act 2024 s 30 obligation as a criminal offence — it is a civil penalty only ($99,000 body corporate / $19,800 individual). Conversely, AI may omit the obligation entirely, causing a client to miss a mandatory report with a 72-hour clock. Missing the SOCI Act reporting stream: AI defaults to describing the NDB 30-day window and (sometimes) the Cyber Security Act 72-hour ransomware report, but routinely omits the SOCI Act's 12-hour (critical incidents) and 72-hour (other reportable incidents) reporting obligations to the ACSC — a third parallel stream with its own civil penalties that applies to a significant number of corporate and infrastructure clients. Privilege waiver in breach response: AI-generated investigation instructions, forensic retainers or board communications that are not structured for a dominant legal purpose lose privilege. Post Optus (May 2024) and Medibank [2025] FCA 167, the 'dominant purpose' test is strictly applied in Australia. AI cannot assess whether a document satisfies this test — it lacks knowledge of the full context of purpose. Privacy tort scope errors: AI applies the tort to pre-10 June 2025 conduct; mis-states the threshold as negligence rather than intentional or reckless; fails to apply proportionality balancing; omits minor-plaintiff limitation provisions (must commence before age 21) and the court-extension outer limit of 6 years; and treats aggravated damages as available (they are not — only exemplary/punitive in exceptional circumstances).

- Hallucinated OAIC determinations: AI invents plausible-sounding case names and holdings that do not exist. In a regulatory response to the OAIC this causes immediate professional conduct exposure. Since 2024 more than 20 Australian court incidents have been recorded; the VLSB+C varied a solicitor's practising certificate in August 2025 following AI-citation misconduct in family law proceedings.
- Treating proposed law as enacted: AI conflates Tranche 2 Privacy Act reforms (broad small business exemption removal, right to erasure, fair and reasonable test) with the enacted 2024 first-tranche changes. Separately, AI may miss the AML/CTF Tranche 2 on-ramp from 1 July 2026 that catches lawyers and other professional services providers regardless of turnover.
- Missing the ransomware reporting penalty category: AI may describe the Cyber Security Act 2024 s 30 obligation as a criminal offence — it is a civil penalty only ($99,000 body corporate / $19,800 individual). Conversely, AI may omit the obligation entirely, causing a client to miss a mandatory report with a 72-hour clock.
- Missing the SOCI Act reporting stream: AI defaults to describing the NDB 30-day window and (sometimes) the Cyber Security Act 72-hour ransomware report, but routinely omits the SOCI Act's 12-hour (critical incidents) and 72-hour (other reportable incidents) reporting obligations to the ACSC — a third parallel stream with its own civil penalties that applies to a significant number of corporate and infrastructure clients.
- Privilege waiver in breach response: AI-generated investigation instructions, forensic retainers or board communications that are not structured for a dominant legal purpose lose privilege. Post Optus (May 2024) and Medibank [2025] FCA 167, the 'dominant purpose' test is strictly applied in Australia. AI cannot assess whether a document satisfies this test — it lacks knowledge of the full context of purpose.
- Privacy tort scope errors: AI applies the tort to pre-10 June 2025 conduct; mis-states the threshold as negligence rather than intentional or reckless; fails to apply proportionality balancing; omits minor-plaintiff limitation provisions (must commence before age 21) and the court-extension outer limit of 6 years; and treats aggravated damages as available (they are not — only exemplary/punitive in exceptional circumstances).

## Review checklist

The lawyer's review checklist

## Review checklist

Every statutory provision or court authority cited by AI has been independently verified on legislation.gov.au and austlii.edu.au — confirm the provision exists, has the correct text, and has not been amended by the 2024 Act or subsequent instruments; check the Australian Clinical Labs [2025] FCA 1224 penalty judgment is correctly described if cited. Every OAIC guideline, determination, or guidance document referenced has been verified at oaic.gov.au — AI hallucination of OAIC determinations is a documented professional conduct risk (VLSB+C action, August 2025); more than 20 Australian court incidents reported since 2024. The commencement date of each obligation has been confirmed against legislation.gov.au: 10 December 2024 (first-tranche 2024 Act); 10 June 2025 (privacy tort); 1 July 2025 (APRA CPS 230); 30 May 2025 (Cyber Security Act ransomware reporting); 1 July 2026 (AML/CTF Privacy Act on-ramp for lawyers and other reporting entities); 10 December 2026 (ADM transparency) — advice must correctly characterise what applies NOW versus what is prospective. All three parallel breach-reporting streams have been checked against the facts: (a) NDB 30-day assessment to OAIC; (b) Cyber Security Act 72-hour ransomware payment report to ASD — civil penalty only, $99,000 body corporate; (c) SOCI Act 12/72-hour incident report to ACSC if client holds critical infrastructure assets. A checklist confirming inapplicability of each stream must be on the file. Client personal data has NOT been entered into any external or public AI tool to generate the output — if actual personal information was used in the prompt, treat as a potential APP 11 breach, assess against the NDB threshold, and check whether state health privacy legislation applies. Privilege strategy is confirmed before any investigation or response document is created: retainer placed by legal counsel (not IT or risk), reports directed to dominant legal purpose, no disclosure to board/PR/regulator that could waive privilege — apply the Optus (dominant purpose) and Medibank [2025] FCA 167 (direct third-party engagement) tests explicitly. The 'which law governs' analysis is lawyer-derived: Commonwealth Privacy Act, applicable state health privacy legislation, APRA prudential standards, Cyber Security Act 2024, SOCI Act, and state surveillance legislation — the applicability of each must be confirmed on the specific facts, not assumed from the AI's default (which defaults to Commonwealth Privacy Act only). For advice involving ADM or algorithmic tools: anti-discrimination law obligations (Sex Discrimination Act, Racial Discrimination Act, Age Discrimination Act, Disability Discrimination Act and state equivalents) have been addressed independently of the Privacy Act ADM provisions, which do not commence until December 2026.

- Every statutory provision or court authority cited by AI has been independently verified on legislation.gov.au and austlii.edu.au — confirm the provision exists, has the correct text, and has not been amended by the 2024 Act or subsequent instruments; check the Australian Clinical Labs [2025] FCA 1224 penalty judgment is correctly described if cited.
- Every OAIC guideline, determination, or guidance document referenced has been verified at oaic.gov.au — AI hallucination of OAIC determinations is a documented professional conduct risk (VLSB+C action, August 2025); more than 20 Australian court incidents reported since 2024.
- The commencement date of each obligation has been confirmed against legislation.gov.au: 10 December 2024 (first-tranche 2024 Act); 10 June 2025 (privacy tort); 1 July 2025 (APRA CPS 230); 30 May 2025 (Cyber Security Act ransomware reporting); 1 July 2026 (AML/CTF Privacy Act on-ramp for lawyers and other reporting entities); 10 December 2026 (ADM transparency) — advice must correctly characterise what applies NOW versus what is prospective.
- All three parallel breach-reporting streams have been checked against the facts: (a) NDB 30-day assessment to OAIC; (b) Cyber Security Act 72-hour ransomware payment report to ASD — civil penalty only, $99,000 body corporate; (c) SOCI Act 12/72-hour incident report to ACSC if client holds critical infrastructure assets. A checklist confirming inapplicability of each stream must be on the file.
- Client personal data has NOT been entered into any external or public AI tool to generate the output — if actual personal information was used in the prompt, treat as a potential APP 11 breach, assess against the NDB threshold, and check whether state health privacy legislation applies.
- Privilege strategy is confirmed before any investigation or response document is created: retainer placed by legal counsel (not IT or risk), reports directed to dominant legal purpose, no disclosure to board/PR/regulator that could waive privilege — apply the Optus (dominant purpose) and Medibank [2025] FCA 167 (direct third-party engagement) tests explicitly.
- The 'which law governs' analysis is lawyer-derived: Commonwealth Privacy Act, applicable state health privacy legislation, APRA prudential standards, Cyber Security Act 2024, SOCI Act, and state surveillance legislation — the applicability of each must be confirmed on the specific facts, not assumed from the AI's default (which defaults to Commonwealth Privacy Act only).
- For advice involving ADM or algorithmic tools: anti-discrimination law obligations (Sex Discrimination Act, Racial Discrimination Act, Age Discrimination Act, Disability Discrimination Act and state equivalents) have been addressed independently of the Privacy Act ADM provisions, which do not commence until December 2026.

## Tasks (32)

| Task | Level | How AI is used | What the lawyer must review |
| --- | --- | --- | --- |
| [Build a privacy-law gap table against the principles that apply](https://ailawframework.com/practice/privacy-data-protection-and-cyber/build-a-privacy-law-gap-table-against-the-principles-that-apply.md) | Yellow | Map the client's described practices and policies to the privacy principles the lawyer has listed, and mark missing notices, security measures and retention… | Confirm whether the client is in scope of the statute Do not treat a template gap as a finding until the facts are verified |
| Draft or review a privacy policy and collection notice | Yellow | Produce a first-draft policy or notice mapped to the required disclosure headings, using only the processing facts the client has supplied. | Delete claims the client cannot stand behind Match automated-decision, overseas-disclosure and retention statements to actual practice |
| [Assess a suspected data breach and draft the notification pack](https://ailawframework.com/practice/privacy-data-protection-and-cyber/assess-a-suspected-data-breach-and-draft-the-notification-pack.md) | Red | The statement to the regulator and the notice to individuals are the lawyer's to write and the tool writes no part of either.… | Decide whether the incident is notifiable under the current statute. Deadlines and 'serious harm' are lawyer calls, and so are the notification route and… |
| Draft a data-breach incident response plan | Yellow | Produce a plan covering detection, containment, assessment, notification and post-incident review, with a RACI and template notices. | Align the plan to every reporting stream that applies Include the privacy regulator, cyber authority, insurers and sector rules Do not rely on one 30-day clock |
| [Build a vendor AI and subprocessors risk questionnaire](https://ailawframework.com/practice/privacy-data-protection-and-cyber/build-a-vendor-ai-and-subprocessors-risk-questionnaire.md) | Yellow | Generate a questionnaire covering residency, subprocessors, training on client data, retention, deletion and audit rights, and score the returned answers into… | Read the vendor's current terms, not the sales deck Decide whether an overseas disclosure is occurring |
| [Draft a privacy or data-protection impact assessment](https://ailawframework.com/practice/privacy-data-protection-and-cyber/draft-a-privacy-or-data-protection-impact-assessment.md) | Yellow | Structure a PIA or DPIA from the project description: information flows, risks, mitigations and sign-off questions. | Check the flow map matches the actual architecture Do not sign off a generic template |
| Build a privacy-tort and misuse-of-information issues table | Yellow | From the facts, list candidate elements of a privacy tort, misuse of information or statutory interference claim, with the documents that support each row. | Apply the cause of action that exists in the governing law Do not import a neighbouring country's new tort |
| [Map automated decisions that use personal information](https://ailawframework.com/practice/privacy-data-protection-and-cyber/map-automated-decisions-that-use-personal-information.md) | Yellow | From the system description, list decisions that use personal information, whether a human reviews them, and what the privacy notice currently says. | Decide what transparency, access or human-review duties apply Check the current instrument; commencement dates change |
| [Build a cross-border transfer mechanism table](https://ailawframework.com/practice/privacy-data-protection-and-cyber/build-a-cross-border-transfer-mechanism-table.md) | Yellow | List destinations, recipients, transfer tools (clauses, adequacy, other) and residual risks from the DPA and architecture papers. | Apply the transfer rule of the governing privacy law Do not mix GDPR adequacy lists with another statute's test |
| [Draft a response pack for a privacy-regulator investigation](https://ailawframework.com/practice/privacy-data-protection-and-cyber/draft-a-response-pack-for-a-privacy-regulator-investigation.md) | Red | From the Commissioner's notice and the incident file, the tool schedules every question, document category and due date the notice sets, and maps the file's… | Read the draft response against the notice question by question: confirm nothing concedes a serious or repeated interference with privacy, or any other… |
| [Map employee-monitoring and workplace-surveillance duties](https://ailawframework.com/practice/privacy-data-protection-and-cyber/map-employee-monitoring-and-workplace-surveillance-duties.md) | Yellow | From the proposed monitoring description, list the surveillance, privacy and employment rules the lawyer has named and the notice or consent steps already… | Check the state surveillance device statute that applies; they differ Do not extend employee-records exemptions to applicants or contractors unless the… |
| [Map critical-infrastructure cyber reporting duties for the asset](https://ailawframework.com/practice/privacy-data-protection-and-cyber/map-critical-infrastructure-cyber-reporting-duties-for-the-asset.md) | Yellow | From the asset description, line the client's systems up against the infrastructure classes and reporting clocks the lawyer has listed, and mark gaps. | Confirm the asset is in the regime before you draft a report Check the class definitions; they are technical |
| Build a record of processing from the interviews and system list | Yellow | Turn the interview notes and system inventory into a processing record: purposes, categories, recipients, locations, retention and security measures. | Check each row with the business owner Watch for invented purposes and default 'global' locations |
| Assemble an access, deletion or correction response pack | Yellow | From the request and the systems list, draft a search log, a proposed response and a list of items that may be withheld, with reasons to check. | Confirm identity, scope and exemptions Do not release another person's information or privileged material |
| Draft a data-processing agreement from the processing facts | Yellow | Produce a first-draft DPA covering instructions, security, subprocessors, deletion and audit from the processing description and the client's precedent. | Match the DPA to the actual processing Check controller and processor labels; they are easy to get backwards |
| Build a 48-hour incident chronology from the tickets and emails | Yellow | Extract times, systems, actors and decisions from the incident tickets and correspondence into a chronology, marking gaps. | Verify times against the source systems Use the chronology for the lawyer's assessment, not as a notification |
| [Run a privacy impact assessment before a client deploys a commercial AI product](https://ailawframework.com/practice/privacy-data-protection-and-cyber/run-a-privacy-impact-assessment-before-a-client-deploys-a-commercial-ai.md) | Yellow | Draft a PIA structure from the vendor documents and data flows: personal information in, outputs, APP 1, 5, 6, 8, 10 and 11 touchpoints. | Confirm what personal and sensitive information actually flows; decide whether the use is necessary and lawful. |
| [Audit a client privacy policy for AI disclosures and chatbot labelling](https://ailawframework.com/practice/privacy-data-protection-and-cyber/audit-a-client-privacy-policy-for-ai-disclosures-and-chatbot-labelling.md) | Green | Compare the policy and collection notices to the client's AI uses and list gaps, including unlabelled public-facing chatbots. | Check each gap against the actual system and approve wording. |
| [Assess whether a client may use held personal information to train or fine-tune a model](https://ailawframework.com/practice/privacy-data-protection-and-cyber/assess-whether-a-client-may-use-held-personal-information-to-train-or-fi.md) | Red | Organise the original collection notices and purposes into a primary/secondary-purpose table for APP 6 analysis. | Decide reasonable expectations and whether consent or opt-out is needed. |
| [Prepare an ADM disclosure for a privacy policy by 10 Dec 2026](https://ailawframework.com/practice/privacy-data-protection-and-cyber/prepare-an-adm-disclosure-for-a-privacy-policy-by-10-dec-2026.md) | Yellow | Identify automated decisions and draft disclosure. | Lawyer confirms against APP 1. |
| [Conduct AI product due diligence for privacy](https://ailawframework.com/practice/privacy-data-protection-and-cyber/conduct-ai-product-due-diligence-for-privacy.md) | Yellow | Assess a commercial AI product's data handling. | Ongoing review, not set-and-forget. |
| [Advise on client-data exposure from consumer AI tools](https://ailawframework.com/practice/privacy-data-protection-and-cyber/advise-on-client-data-exposure-from-consumer-ai-tools.md) | Yellow | Assess exposure and remediation. | Lawyer settles. |
| [Update collection notices for AI chatbots](https://ailawframework.com/practice/privacy-data-protection-and-cyber/update-collection-notices-for-ai-chatbots.md) | Green | Draft chatbot notices. | Check APP 5. |
| [Build an AI privacy impact assessment](https://ailawframework.com/practice/privacy-data-protection-and-cyber/build-an-ai-privacy-impact-assessment.md) | Yellow | Draft PIA for an AI system. | Lawyer confirms. |
| Assess privacy compliance of a commercially available AI product | Yellow | Apply OAIC guidance to the product terms. | Lawyer advises. |
| Draft a law firm policy against putting client data into free AI tools | Green | Draft rules on data security. | Principal approves. |
| Draft a notifiable data breach assessment involving AI tools | Red | Assess serious harm. | Lawyer settles. |
| Draft the statement to the OAIC about a notifiable data breach from the incident log | Yellow | Draft the statement using only facts written in the incident log. | Match every affected group and every step to the log and to the OAIC notifiable data breaches page. |
| Compare a privacy-policy draft with the OAIC's APP 1 chapter | Yellow | List policy clauses beside the headings in the APP 1 chapter. | Open the APP 1 chapter and decide which clauses meet it. Do not rely on a generated summary of the chapter. |
| List the personal information a client-intake chatbot would collect and check it against the Privacy Act 1988 | Yellow | List the fields the chatbot asks for. | Read the Privacy Act compilation before advising whether the collection is allowed. The list is not that advice. |
| Prepare the file note deciding whether an AI-tool incident is a notifiable data breach | Red | Set out the incident facts from the log. Do not decide whether notification is required. | The lawyer decides, using the incident log and the OAIC notifiable data breaches page. |
| Draft the transparency paragraph of a privacy policy and check it against the OAIC APP 1 chapter | Yellow | Draft a transparency paragraph from the client's description of the tool. | Check the paragraph against APP 1 as the chapter explains it, and against what the tool actually does. |

## Jurisdiction notes

### Australia

#### Legal sources

Privacy Act 1988 (Cth) and its 2024 amendments, the APPs, the Notifiable Data Breaches scheme, the Cyber Security Act 2024 (Cth) and the Security of Critical Infrastructure Act 2018 (Cth).

#### Moving law

The regime is layered and changing fast, so an invented authority flows straight into compliance advice.

#### Client data

Putting client data into an AI tool can itself be the breach the firm is advising on. Treat tool use as a confidentiality and governance question.

### New Zealand

#### Governing law

Privacy Act 2020 and the information privacy principles, not the APPs. IPP 12 covers disclosure outside New Zealand.

#### IPP 3A

Notification of indirect collection, in force since 1 May 2026.

#### Breaches

If a breach has caused or is likely to cause serious harm, notify the Privacy Commissioner and affected people as soon as practically able. Not the OAIC NDB scheme.

#### Commissioner's AI expectations

Privacy impact assessment before use; human review before acting on output; no personal input unless the provider confirms it will not retain or disclose it.

- Do not enter personal information into a public generative tool unless the provider has confirmed it will not retain or disclose the input.

Sources:

- [OPC: Privacy Act 2020 information privacy principles](https://www.privacy.org.nz/privacy-principles/) (checked 2026-09-23)
- [OPC: Artificial Intelligence and the Information Privacy Principles](https://www.privacy.org.nz/resources-and-learning/a-z-topics/ai/) (checked 2026-09-23)
- [OPC: Generative artificial intelligence (15 June 2023)](https://www.privacy.org.nz/resources-and-learning/a-z-topics/ai/generative-artificial-intelligence/) (checked 2026-09-23)
- [OPC: IPP 3A guidance](https://www.privacy.org.nz/resources-and-learning/a-z-topics/ipp3a/) (checked 2026-09-23)
- [OPC: sorting out privacy breaches](https://www.privacy.org.nz/responsibilities/privacy-breaches/) (checked 2026-09-23)

### United States

Local name: Privacy, data protection and cyber

#### No general statute

No federal equivalent of the Privacy Act 1988 (Cth). Start with Rule 1.6, then the sector statute that covers the data.

#### Sector statutes

HIPAA for PHI; GLBA for non-public financial data; CCPA as amended by the CPRA (know, delete, correct, opt out of sale or sharing, limit sensitive data).

#### Colorado

SB26-189 (signed 14 May 2026) repeals and re-enacts SB24-205; ADMT duties from 1 January 2027. HB26-1263 (signed 1 July 2026) adds chatbot duties from the same date.

#### Vendor AI

Formal Opinion 512: understand retention, training use and access before you input client information.

- Do not assume ‘US privacy law’ is one instrument. Name the statute that covers the data.

Sources:

- [ABA Model Rule 1.6, Confidentiality of Information](https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/rule_1_6_confidentiality_of_information/) (checked 2026-09-23)
- [ABA Formal Opinion 512, Generative Artificial Intelligence Tools (29 July 2024)](https://www.americanbar.org/content/dam/aba/administrative/professional_responsibility/ethics-opinions/aba-formal-opinion-512.pdf) (checked 2026-09-23)
- [California Attorney General: California Consumer Privacy Act (CCPA)](https://oag.ca.gov/privacy/ccpa) (checked 2026-09-23)
- [HHS: HIPAA covered entities and business associates](https://www.hhs.gov/hipaa/for-professionals/covered-entities/index.html) (checked 2026-09-23)
- [Colorado Attorney General: Automated Decision-Making Technology and Chatbot Safety rulemaking](https://coag.gov/ai/) (checked 2026-09-23)
- [Colorado Senate Bill 26-189 (became law 14 May 2026)](https://leg.colorado.gov/bills/sb26-189) (checked 2026-09-23)

*Practice guidance for planning and review. Confirm current primary sources and the requirements of the actual matter. This is general information, not legal advice.*