You're offline, showing the last version we saved.
Free template
AI incident response plan: the first 72 hours
Most AI incidents start small: a client file pasted into a personal account, a connector that could see too much, a voice note that was not really the client. A plan agreed in advance turns panic into a checklist. The 72 hours here is an internal target to get the facts straight and decide on notifications, not a legal deadline.
Updated · Editable Word document (.docx) · Free with a firm profile
Free with a firm profile
Get the editable AI incident response plan: the first 72 hoursDownload the AI incident response plan: the first 72 hours
The Word document (.docx), ready to fill in. Your profile unlocks every template on The Firm.
Or build it with AI
Open Claude or ChatGPT with the prompt filled in. It writes the finished document for your firm straight away, then asks a few questions to tailor it. No client data needed.
Works without client data. Don't paste client names, tax file numbers or ledgers into an AI tool your firm hasn't approved for them. AI data safety playbook.
How to use it
How to use this template
1
Name the people first
Fill in the contacts table before anything happens. In a small firm, the AI lead and the privacy lead may be the same person.
2
Make owning up safe
Tell the team they will not be in trouble for reporting a mistake quickly. Slow reporting is what turns a small incident into a big one. Shadow AI: find the AI your team already uses explains how to run a no-blame amnesty.
3
Keep it with your AI policy
Store it next to your AI policy and data breach response plan, and make sure everyone knows where it is.
4
Run a short practice
Once a year, walk through a made-up example, such as a client file pasted into a personal chat, and update the plan with what you learn.
The template
The template in full
[Firm name] AI incident response plan
Replace everything in [square brackets].
1. Who is involved
Role
Name and contact
What they do
AI lead
[ ]
Runs the response and keeps the incident log.
Privacy lead
[ ]
Assesses whether personal information is involved and whether it is a notifiable data breach.
IT provider
[ ]
Revokes access, checks devices and connectors, and preserves records.
Insurer or broker
[ ]
Told early: many policies require prompt notice.
AI vendor support
[ ]
Asked about retention, deletion and logs.
2. The first hour: stop and contain
Stop using the tool or connector involved.
Tell the AI lead straight away.
If money may have moved, call the bank now.
Disconnect or revoke the connector, account or access involved.
Delete the chat or file from the tool if possible, but first note what was shared, when and where.
Change any passwords that may have been exposed.
3. The first 24 hours: get the facts
Which clients and what information were involved?
Which tool, which plan, and does that plan train on data or let staff review it?
How long does the vendor keep the data, and can it be deleted? Ask the vendor in writing.
Who else could have seen it?
Tell the insurer or broker.
For a scam: report to ReportCyber at cyber.gov.au and to Scamwatch.
4. By 72 hours: decide and tell
The privacy lead assesses whether it is likely to result in serious harm to anyone.
If it is an eligible data breach, notify the people affected and the Office of the Australian Information Commissioner as soon as practicable.
Decide what to tell affected clients, even if notification is not required.
If clients outside Australia are involved, check their local rules: some set their own deadlines.
5. Afterwards: fix the cause
Fix what allowed it: settings, access, the approved tool or the process.
Update the AI policy and the AI tools register.
Run a short refresher with the team, without blame.
Close the incident log with what changed.
6. Incident log
Date and time
What happened
Who
Action taken
[ ]
[ ]
[ ]
[ ]
[ ]
[ ]
[ ]
[ ]
[ ]
[ ]
[ ]
[ ]
This template is a starting point, not legal advice. Your obligations depend on where you practise, your professional body and your clients; take advice on anything you are unsure of.
Common questions
Frequently asked questions
What counts as an AI incident?
Client data going into an unapproved or personal AI tool, a connector reaching or changing more than it should, AI output sent to a client without review that caused harm, or a scam using AI such as a cloned voice or deepfake call. AI scams and deepfakes covers cloned voices, deepfake calls and the safe phrase.
Is 72 hours a legal deadline in Australia?
No. It is an internal target to get the facts straight and decide on notifications. Under the Notifiable Data Breaches scheme, you assess a suspected breach promptly and notify as soon as practicable if it is eligible. Clients in other countries may be covered by their own rules.
Is pasting client data into ChatGPT a data breach?
It may be an unauthorised disclosure. Whether it is a notifiable breach depends on the information involved and whether it is likely to result in serious harm, so treat it as a possible incident and follow this plan. Client data and AI covers what can go into an AI tool in the first place.