Cybersecurity Awareness · AI & Privacy

AI tools and data privacy: beware before you share

21 Apr 20253 MIN READUPDATED 15 NOV 2026

AI chatbots and assistants are powerful and convenient, but anything you paste into them may be stored, reviewed or reused. Here are the risks and simple rules for using AI safely at work.

AI tools such as ChatGPT, Microsoft Copilot, Google Gemini and DeepSeek are changing how we work. They can summarise reports, draft emails and write code in seconds, and many are free. But before you paste a confidential document into a chatbot or upload files to an AI service, stop and think about where that data goes.

What are the risks?

1. Your data may be stored and used for training

Some AI services, especially free consumer versions, keep what you type and may use it to improve their models. Conversations may also be reviewed by the provider’s staff. If your input includes customer records, financial figures, source code or internal plans, you may be handing that data to a third party.

2. It is not always clear how data is handled

Many services do not make it clear how long they keep data, who can see it, where it is stored or how it is protected. Without a proper business agreement, you cannot be sure your information is safe.

3. You lose control once data is shared

Once uploaded, data can travel to the provider’s systems, its subcontractors and its backups. It could be exposed in a breach at the provider, or disclosed under a legal order.

4. “Delete” may not mean deleted

Deleting a conversation from your history does not guarantee that the data has been removed from the provider’s systems or backups.

5. Compliance risks

Sharing personal data with an unapproved AI service may breach Malaysia’s Personal Data Protection Act (PDPA), including its rules on transferring personal data outside Malaysia, as well as customer contracts and company policy. The result can be penalties and reputational damage.

Watch: beware before you share with AI

A real-world warning

In 2023, Samsung engineers were reported to have pasted confidential source code and internal meeting notes into ChatGPT while asking for help. Once submitted, the company had no way to get the data back, and it soon restricted the use of generative AI tools on company devices.

How to use AI safely

  • Assume anything you share could be seen by others. If you would not post it publicly, do not paste it into a public AI tool.
  • Never share confidential documents, client information, personal data, passwords, internal strategy or source code with unapproved tools.
  • Use company-approved AI tools that have been through a security review, ideally business versions that do not use your data for training.
  • Remove identifying details before asking for help, for example by replacing names and figures with placeholders.
  • Check privacy settings and terms, and turn off chat history or training where the tool allows it.
  • Be careful with permissions. Do not connect AI assistants to your email, files or calendar unless your organisation has approved it.
  • Check the output. AI can be confidently wrong, so review its answers before relying on them.
  • Report accidental exposure to your security team straight away.

Watch: the dangers of AI for data privacy

For organisations

Banning AI rarely works; staff simply use it on personal devices. A better approach is to provide approved tools, publish a clear AI use policy, train staff on it and use data loss prevention to stop sensitive data being pasted into unapproved services.

AI should help, not hurt. Be smart about what you share and where you share it.

How NetWitz helpsWe help organisations adopt AI safely, with approved tools, usage policies and data loss prevention controls, through our data, identity and change protection service.

Get the next guide in your inbox.

Join the NetWitz Cybersecurity Newsletter and Awareness Campaign. Unsubscribe at any time.

More Cybersecurity Awareness

Want to protect your organisation, not just yourself?

Talk to us about awareness training, managed security and the controls behind these guides.

NetWitz