Cases
AI data breach examples and lessons
Four well documented cases, each with a different cause. None needed a sophisticated attacker.
Last reviewed against legislation.gov.uk, National Cyber Security Centre and OWASP GenAI Security Project. Sources
The cases
Cause and lesson
Each case is told from the account of whoever found or disclosed it.
ChatGPT, March 2023
A flaw in a third-party code library led to some users seeing other users' chat titles. Payment-related details also showed for roughly 1.2% of Plus subscribers who used the service during a nine-hour stretch [9]. Lesson: your supplier's own dependencies are part of your risk.
Samsung, April 2023
Engineers put internal source code into ChatGPT, according to a company memo that the press reported in May 2023 [10]. Lesson: staff behaviour can be the exposure route, with no attacker involved.
Microsoft AI research, 2023
A storage link published in a code repository opened up about 38 TB, among it backups of two employees' workstations and over 30,000 internal Teams messages [8]. Lesson: one over-broad link can open an entire account.
DeepSeek, January 2025
An unprotected database held more than a million log lines, with conversations, secret keys and operational details [7]. Lesson: logs are data stores and need the same controls as any other.
Pattern
What they share
Wiz said it limited its DeepSeek work to listing what was reachable and ran no intrusive queries [7], and it reported each finding to the company concerned [7] [8]. In every case the remedy was to close access or change a process, which points to basic controls mattering as much as anything specific to AI.
The wider picture
Controls in short
IBM's 2025 research found that, among breached organisations with an AI-related incident, 97% had not put suitable access controls around their AI systems [6]. It also reported that 13% of those surveyed had suffered an attack touching their AI systems [6].
Use these as scenarios for your response plan, and see the full timeline.
Read the sources, not summaries
The accounts above come from the people who found the problems and from press reports. Where a company has not published its own account, treat the details as reported rather than confirmed.
Keep reading
Read next
What to do
A practical order of work for UK organisations after data leaks through an AI tool: awareness, containment, scoping, reporting, supplier and plan.
Read the guide
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 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.