Dutch City Uploads Thousands of Personal Files to ChatGPT — What This Means for Your Privacy

Disclosure: This article contains affiliate links. We may earn a commission at no extra cost to you.

December 2025: The city of Eindhoven, Netherlands, reports a data breach. Employees uploaded 2,368 files containing personal data to public AI tools like ChatGPT and Claude. In just 30 days.

The files included youth welfare documents, job applicant CVs, and reflection reports about vulnerable citizens. The Dutch Data Protection Authority (Autoriteit Persoonsgegevens) classified it as a data breach. And here's the part that should worry everyone: the mayor defended the practice.

This isn't just a story about one city's mistake. It reveals a fundamental problem with how organizations use AI tools. And it directly affects what happens to your personal data when you interact with any company, government, or service provider.

What Happened in Eindhoven

The breach was discovered during a sample analysis of file uploads from September 23 to October 23, 2025. City employees had been using public AI chatbots like ChatGPT and Claude to help with their work. Sounds innocent, right?

The problem: they were uploading files containing personal information. Not just names and addresses, but sensitive documents about vulnerable citizens.

The Leaked Data Included:

• Youth Welfare Act documents (Jeugdwet) with details about children and families receiving care
• Reflection reports containing assessments of citizens with financial or personal problems
• CVs of job applicants with names, addresses, phone numbers, and work history
• Photos of citizens attached to case files
• Internal memos about individual cases

The city couldn't even determine the full scope. Why? Because the logs only go back 30 days. Files uploaded before September 23 are simply unknown. The breach could be much larger.

The AI Training Problem

Here's what makes this breach different from a traditional hack. The data wasn't stolen by criminals. It was voluntarily given to AI companies. And those companies may use it to train their models.

OpenAI (ChatGPT) Is Clear About This

From OpenAI's own FAQ:

"We may use content submitted to ChatGPT and our other services for individuals to improve model performance. For example, depending on a user's settings, we may use the user's prompts, the model's responses, and other content such as images and files to improve model performance."

That's not ambiguous. Files you upload can be used to train ChatGPT. Unless you specifically opt out (which most people don't do, and which city employees certainly didn't), your data becomes training material.

Claude (Anthropic) Has Similar Policies

From Claude's privacy documentation:

"You choose to allow us to use your chats and coding sessions to improve Claude, [or] your conversations are flagged for safety review (in which case we may use or analyze them to improve our ability to detect and enforce our Usage Policy)."

Again: your conversations and uploaded files can be used for AI training and safety reviews.

Microsoft Copilot Is Different

Here's the contrast. Microsoft's enterprise Copilot (the secured version Eindhoven switched to after the breach) operates differently:

• Files are not opened by Microsoft
• Content is not used to train their models
• At most, they compare image hashes for safety (without viewing content)

This is why organizations need enterprise AI tools with proper data processing agreements, not free public chatbots.

The Political Response: "This Is Actually Positive"

You might expect city leadership to be alarmed. Instead, Mayor Jeroen Dijsselbloem (yes, the former Dutch Finance Minister and Eurogroup President) wrote to the city council:

"The use of AI offers opportunities to do our work more efficiently. From this perspective, it's positive that employees see opportunities and engage with AI. Initial assessment suggests employees used it to improve municipal tasks and services to residents."

Read that again. Citizens' personal data leaked to AI training systems, and the official response is that it's "positive" employees are experimenting with AI.

The city also downplayed the risk. From their statement:

"According to experts we consulted, the risk is limited compared to, for example, a data leak where data is stolen. Individual data uploaded to an AI tool cannot easily be extracted by third parties and misused."

This misses the point entirely. The concern isn't that hackers will query ChatGPT to get your data back. It's that your personal information becomes permanently embedded in AI models that will be used by millions of people worldwide.

The Ironic Timing

Here's what makes this story even more remarkable. Eindhoven was under enhanced supervision by the Dutch Data Protection Authority from March 2023 until March 2025. For two full years.

Why? Because they had a pattern of:

• Collecting too much personal data
• Storing data longer than legally allowed
• Reporting data breaches late or not at all

The enhanced supervision ended in March 2025. Seven months later, this breach was discovered. Some habits are hard to break.

If a Government Does This, What About Your Employer?

This is the question you should be asking. Eindhoven is a Dutch city subject to GDPR, with a privacy officer, with legal obligations, and under recent scrutiny from the data protection authority. They still failed.

Now consider:

• Your employer's HR department processing your performance reviews
• Your doctor's office handling your medical records
• Your bank's customer service analyzing your complaint
• Your child's school managing student records

How many employees at these organizations use ChatGPT? How many paste sensitive information into AI tools to "help with their work"? How many have data processing agreements in place?

The answer, for most organizations, is: lots of employees use AI, and almost none have proper controls.

How Can You Protect Yourself?

You can't control what organizations do with your data. But you can minimize your exposure and make informed choices.

1. Assume Your Data Gets Shared

Any information you provide to any organization may end up in an AI system. Act accordingly:

• Provide only the minimum information required
• Ask how your data will be processed (especially for sensitive matters)
• Request that your data be handled without AI tools when possible

2. Protect Your Digital Identity

Your IP address links your online activities to your location. Hide it when possible.

Verify your VPN is working:

• Check your IP at myip.foo (should show VPN location)
• Run our DNS Leak Test
• Run our WebRTC Leak Test

3. Be Careful What You Share With AI Tools Yourself

If you use ChatGPT, Claude, or similar tools:

• Never upload files with personal information about others
• Opt out of training data collection in settings
• Use enterprise versions for work (with proper agreements)
• Assume everything you type may be read by humans and used for training

4. Use Browser Privacy Tools

Block trackers that collect data about your browsing:

• uBlock Origin: Blocks ads and trackers
• Privacy Badger: Learns and blocks invisible trackers
• Our WebRTC Blocker: Prevents IP leaks (free download)

See our complete guide: 10 Browser Extensions That Protect Your Privacy.

5. Know Your Rights Under GDPR

If you're in the EU, you have rights:

• Right to access: Ask what data an organization holds about you
• Right to erasure: Request deletion of your data
• Right to object: Refuse certain types of processing
• Right to complain: Report violations to your data protection authority

For Dutch residents affected by the Eindhoven breach: you can file a complaint with the Autoriteit Persoonsgegevens.

Common Questions

Can I find out if my data was uploaded to ChatGPT?

Unfortunately, no. There's no way to query AI models to see if they were trained on your data. The city of Eindhoven itself doesn't know whose data was uploaded. This is part of the problem.

Is my data "in" ChatGPT now forever?

Not exactly "in" it like a database. AI training works differently. Your data influenced the model's patterns during training, but can't be retrieved directly. However, models can sometimes reproduce training data in outputs, which is a privacy concern researchers are actively studying.

Didn't the city ask OpenAI to delete the data?

Yes, Eindhoven requested deletion. But there's a problem: if the data was already used in a training run, it's too late. The model has already learned from it. You can delete the uploaded files, but you can't "untrain" a model.

Why doesn't GDPR prevent this?

GDPR does prohibit this. That's why it's classified as a data breach. But GDPR is enforced after the fact. By the time violations are discovered and fined, your data has already been processed. Prevention depends on organizations following the rules, which many don't.

Should I stop using AI tools entirely?

Not necessarily. Just be smart about it. Never input others' personal data. Opt out of training data collection. Use enterprise tools for sensitive work. And assume anything you type could be seen by humans reviewing the system.

Conclusion

The Eindhoven data breach reveals an uncomfortable truth: the AI revolution is happening faster than privacy protections can keep up. Government employees, trying to work more efficiently, accidentally fed citizens' personal data into global AI training systems. And leadership thinks this is "positive."

Key takeaways:

• Public AI tools like ChatGPT and Claude use your inputs for training
• Organizations are uploading sensitive data without proper controls
• Even GDPR-compliant countries have organizations that fail basic privacy
• Your data can end up in AI systems without your knowledge or consent
• Once data is used for training, it can't be "untrained"

The good news: you can take steps to protect yourself. Use a VPN to hide your location. Use browser extensions to block trackers. Be mindful about what data you share with any organization. And demand better from the companies and governments that handle your information.

If a city government under enhanced privacy supervision can accidentally leak citizens' data to AI companies, imagine what's happening at organizations with less oversight. Your privacy is your responsibility. Start protecting it now.