Actions
What to do now about AI data breaches
Seven actions, in the order most organisations need them. The first four are for the day you find a problem; the last three are for this week, before you do.
Last reviewed against legislation.gov.uk, National Cyber Security Centre and OWASP GenAI Security Project. Sources
How to use this
Two tracks
If you are in an incident now, start with action one and read the first response steps alongside. If you are preparing, work from the bottom up: the plan, the supplier contacts and the written record come first, because they cannot be improvised under pressure.
The 72 hour reporting target is measured from becoming aware of a breach [1]. That is why preparing the decision owner matters more than any tool.
Where to read more
One page per action
Each action links to a page that goes further.
Judgement
What not to do
Do not wait for certainty. Details can reach the regulator in stages [1], and the NCSC expects organisations to have a response plan for AI incidents ready before one happens [3].
Do not delete the evidence
Clearing logs or chat history to tidy up removes what you need to scope the breach and to show your reasoning later.
What to do
Seven steps, in order
-
Step 1
Note when you became aware[1]
Write down the time, who raised it and what they said. The notification target is counted from becoming aware of the breach.
Who: Incident lead and data protection officer
-
Step 2
Stop the exposure and rotate secrets[7]
Close the open service, revoke the shared link or suspend the tool, then replace any key or token that appeared in logs, prompts or repositories. Wiz found secret keys in the exposed DeepSeek logs and a storage token that gave full control.
Who: IT and cloud teams
-
Step 3
Establish what data, whose, and how much[1]
A report must describe the nature of the breach, the categories and approximate numbers of people and records, the likely consequences and the measures taken. Start gathering those facts at once.
Who: Data protection officer and security lead
-
Step 4
Decide on reporting and record why[1]
Notify the regulator unless the breach is unlikely to risk people's rights and freedoms. Late notices must give reasons for the delay, and information can be sent in phases.
Who: Data protection officer and a director
-
Step 5
Tell affected people when the risk is high[2]
Where the breach is likely to result in a high risk, communicate it to individuals without undue delay, in clear and plain language.
Who: Communications lead and data protection officer
-
Step 6
Get the supplier's account in writing[1]
If an AI supplier is involved, ask what happened, what data of yours was affected and what has been fixed. A processor must tell the controller without undue delay after becoming aware.
Who: Procurement and security leads
-
Step 7
Bring AI into the response plan[3]
The NCSC says organisations should have an appropriate response plan for AI-related incidents and that leaders should understand their data, model and software supply chains.
Who: Directors and security leads
Keep reading
Read next
Timeline
A dated, sourced timeline of public AI-related data exposures and the UK guidance that followed, from the ChatGPT bug to the DeepSeek database.
Read the guide
Questions
Sourced answers on reporting an AI data breach, the 72 hour rule, telling people, supplier breaches, leaked keys and what UK organisations should do.
Read the guide
First response
What to do first when data has leaked through an AI tool: record awareness, stop the exposure, replace secrets, scope the data and decide on reporting.
Read the guideHow CyPro can help
Need a hand working through this?
Finding out whether an incident reaches you is one job. Checking your own systems, tightening access and briefing the board are several more. CyPro's responders and consultants can take any of them on.
Incident response
If you may already be compromised, our responders contain it, preserve the evidence, find out what was touched and get you running again safely.
More on cypro.co.uk
24/7 security monitoring
Our SOC analysts watch your systems day and night for the signs of an attack in progress: unexpected processes, new outbound connections, logins that do not fit.
More on cypro.co.uk
Fractional CISO for the board
A senior security lead who advises the board on what to fix first, what to accept and what to tell customers, and owns the decision record.
More on cypro.co.uk
Penetration testing
A focused test of your systems and the controls around them, so you know they hold before an attacker finds out for you.
More on cypro.co.uk
Hardening and segmentation
Our consultants put the controls in place that limit the damage when something is missed: firewall rules, network isolation and tighter admin access.
More on cypro.co.uk
Vulnerability scanning
Find every internet-facing system you run, including the forgotten ones, and see which are still on a vulnerable build.
See the scanning service
Talk to a specialist
Worried this affects you?
When data has leaked through an AI tool, its logs or its storage, the early decisions shape what you must report and to whom. An incident response team helps you contain the exposure, establish what was involved and prepare your notifications.