AI SecurityNov 16, 2026

An Employee Pasted Customer Data Into an AI Tool: A Response Playbook for Small Teams

Sudip Bhandari
Sudip Bhandari
Co-founder, Sequirly
An Employee Pasted Customer Data Into an AI Tool: A Response Playbook for Small Teams

If an employee pasted customer data into an AI tool, the first hour is for facts. Find out which account it went into, what exactly was pasted and whether training was on, then decide within 24 hours whether you have to tell anyone.

Those three facts decide how serious this is. A paste into a company account with training off is a very different problem from a paste into a free personal account with training on.

Honestly, I'd rather you hear about it from your employee than from a customer. So the first thing to protect is the next report. If this one ends in a telling-off, the next person will stay quiet.

Note
Under Article 33 of the GDPR, a company that controls personal data has to report a breach to the regulator without undue delay and, where feasible, within 72 hours of becoming aware of it, unless the breach is unlikely to put people at risk. The clock starts when you become aware, which is why the first hour matters.

Let's start with the three questions that tell you what you're dealing with.

Start with three questions

QuestionWhy it mattersWhere to look
Which account was it?Company accounts on a business plan come with contract terms. Personal accounts don't.Ask the employee, then check your workspace's member list
What exactly was pasted?Names and emails are one level of risk. Payment details, health data and credentials are another.Ask for the chat link or a screenshot
Was training on?If it was, the text may be used to train the model.ChatGPT: Settings, Data controls. Claude: Settings, Privacy

One practical tip: don't ask the employee to paste the data again to show you. The chat link or a screenshot of the conversation is enough.

The first hour

Five steps to take in the first hour after customer data is pasted into an AI tool
1.Thank them, and ask the three questions. Say it plainly: you're glad they told you, and nobody's in trouble for reporting.
2.Write down the facts. Note who it was, when it happened, which tool and account, what was pasted and roughly how many customers it covers. Take a screenshot of the chat before you delete anything.
3.Delete the chat. In ChatGPT, a deleted chat disappears from the history right away, and OpenAI says it removes it from its systems within 30 days, unless it has to keep it for security or legal reasons. Archiving doesn't delete anything. In Claude, a deleted conversation leaves your history immediately and Anthropic's back-end storage within 30 days, and Anthropic says it won't be used for future training.
4.Switch off training. In ChatGPT, turn off "Improve the model for everyone" under Data controls. In Claude, turn off "Help improve Claude" under Privacy. That protects what you send next. It can't pull back anything that was already used for training, and here's how that works for ChatGPT.
5.Rotate anything that works like a password. If the paste included an API key, a token or a password, change it now. Here's the first-hour guide for a pasted API key.

The first day: do you have to tell anyone?

This is the part everyone worries about, so let's take it in order.

Is personal data involved? The GDPR defines a personal data breach to include unauthorised disclosure of personal data. Whether a paste counts depends on the facts, mostly which account it went into and what your contract with the provider says. If you aren't sure, treat it as a possible breach until a lawyer tells you otherwise.

Does a deadline apply? If the GDPR covers you, the 72 hours in the box above is the deadline for telling the regulator. The same law says you must also tell the people affected when the breach is likely to put them at high risk, with exceptions such as data that was encrypted. In the US, the rules depend on the state and the industry.

What do your contracts say? Your customer agreements may set their own notice period, so read the notice clause before you decide.

Have you written it down? The GDPR asks you to record every breach, including the ones you decide not to report. A short log protects you if anyone asks later.

Here's a rough way to sort it.

What happenedUsual next step
Company account, training off, low-sensitivity dataDocument it, fix the cause and move on
Company account, customer personal dataDocument it, check your contract and ask a lawyer whether you need to notify
Personal account or training on, customer personal dataTreat it as a possible breach, call a lawyer and expect to tell the customer
Payment, health or government ID data, or credentialsTreat it as serious in every case, and rotate credentials first

This isn't legal advice, and the rules vary by country and industry. It's a starting point for the conversation with your lawyer.

Sequirly
Limited time · No credit card required

Prevent accidental data leaks to ChatGPT, Claude, and Gemini.

Sequirly scans your prompts and uploaded files before they're sent. If it finds credentials, client records, or API keys, it stops you before the request goes out.

The first week

If you do have to tell a customer, keep the message plain and short.

We found out on [date] that one of our team members entered [type of data] into [tool and account type] on [date]. We deleted the conversation on [date] and turned off the setting that allows model training. The data involved was [description]. We recommend [what the customer should do, if anything]. You can reach [name] at [contact] with questions.

Then fix the cause, because the same paste will happen again if nothing changes.

1.Move the work to company accounts. The risk sits in personal accounts, where your company has no visibility and no contract.
2.Write one rule. Say which data can go into AI tools and which can't. Our policy generator gets you a draft in minutes.
3.Find out what else is in there. Ask your team without blame, and run our free scanner on an exported chat history. It runs in your browser, and nothing is uploaded.
4.Add a check at the moment of the paste. A policy can't stop someone pasting a customer list at 11pm before a deadline.

What to put in your incident log

Keep one row per incident. A shared spreadsheet is enough.

FieldWhat to write
Date and timeWhen it happened and when you found out
Reported byWho told you, and how
Tool and accountFor example, ChatGPT Free, personal account
Data involvedThe type of data and roughly how many people it covers
Training settingOn or off when it happened
Actions takenChat deleted, training turned off, credentials rotated
Who was toldCustomers, regulator, insurer, with dates
Owner of the follow-upOne name, and a date

Where Sequirly fits

A response plan limits the damage after a paste. Sequirly checks what you're about to send to ChatGPT, Claude or Gemini and warns you before sensitive data leaves your browser. The admin sees which category of data was flagged and which tool was involved, and never the content of the prompt.

If you want rules built around the customer data you handle, the Sequirly team will sit down with you, understand your requirements and build the setup around your workflow.

Sequirly has limits too. It works in the browser, so it doesn't cover desktop AI apps, and it doesn't replace a response plan.

Common Questions

What should I do if an employee pasted customer data into ChatGPT?

Find out which account it went into, what was pasted and whether training was on. Then delete the chat, turn off training, rotate any credentials and decide within 24 hours whether you have to notify anyone.

Do I have to report an employee pasting customer data into an AI tool?

It depends on the law that applies to you, the type of data and your customer contracts. Under the GDPR, a personal data breach has to be reported to the regulator within 72 hours where feasible, unless it's unlikely to put people at risk. Ask a lawyer if you aren't sure.

Can OpenAI or Anthropic delete what an employee pasted?

Deleting the chat removes it from the history immediately, and both companies say deleted chats are removed from their systems within 30 days, with some exceptions. Neither can take back data that was already used for training.

Is pasting customer data into ChatGPT a GDPR breach?

It can be. The GDPR defines a personal data breach to include unauthorised disclosure of personal data. Whether a paste counts depends on the account it went into and your contract with the provider.

Should I discipline the employee?

A single honest mistake rarely calls for it. The bigger risk is that people hide the next one, so reward the report and fix the process.

Where to start

Write down the three answers for this incident, then log it. After that, generate a policy so the next person knows the rule.

Then try Sequirly free so the paste gets checked before it's sent. Or if you'd like to talk it through first, book a quick demo.

Start Protecting Your Data

Ready to Prevent AI Data Leaks?

Sequirly catches sensitive data in real-time, before it leaves your browser. Set up in 2 minutes, runs locally, zero training required.

Sequirly is the safety layer of your AI stack. Local scanning, 2-minute setup, free plan.