Updated November 14, 2026
In April 2023, Samsung engineers pasted internal source code into ChatGPT to help check it for errors. Within weeks, Samsung banned generative AI tools on company devices for one of its biggest divisions. Nobody was trying to leak anything. They were just trying to get work done faster.
That story still sums up the main risk of AI at work. The danger usually isn’t hackers. It’s ordinary people pasting the wrong thing into the wrong tool. This guide covers AI privacy at work in plain terms: what happens to what you type, what’s safe to share, and the settings worth checking today.
Key takeaways
- Most AI privacy at work problems come from people pasting the wrong data into the wrong tool.
- Use approved tools for anything internal or confidential.
- Check training, memory and connected-app settings today.
- Anonymise names and figures before you paste whenever you can.
What happens to what you type
When you use an AI assistant, your prompt and any files you upload are sent to the provider’s servers to generate a response. What happens next depends on the plan and settings.
On personal plans, providers may use conversations to improve their models unless you opt out. OpenAI’s pricing page, for example, lists model training as on with an opt-out available for its individual plans, and Anthropic lists an opt-out for Claude’s individual plans too. Business and enterprise plans usually come with stricter terms, admin controls and data retention settings. The details vary, so read the data terms for the exact plan you use.
The three questions to ask before pasting anything
- Is this tool approved by my company? If there’s an approved tool, use it. If there’s a ban, respect it.
- Would it be a problem if this text appeared outside the company? Customer data, financials, unreleased plans, source code and HR matters all fail this test.
- Does this contain someone else’s personal data? Names, emails, phone numbers, health information and ID numbers need extra care and may fall under data protection law.
If any answer worries you, don’t paste it into a personal account.
What’s usually safe and what isn’t
| Usually fine | Use only approved tools | Don’t share |
|---|---|---|
| Public information and general questions | Internal documents and meeting notes | Passwords, API keys and credentials |
| Your own rough drafts with no client details | Customer emails and support tickets | Unreleased financial results |
| Formula and coding questions with dummy data | Source code | Health, ID or bank details of real people |
| Rewriting text you’d post publicly anyway | Spreadsheets with business data | Anything under NDA or legal privilege |
Settings to check today
- Model training: in ChatGPT, Claude and Gemini, find the data or privacy settings and decide whether your conversations can be used for training.
- Memory: many assistants now remember details across chats. Check what’s stored and delete anything work-related you don’t want kept.
- Connected apps: if you’ve linked email, Drive or calendars, review which ones the assistant can read.
- Shared links: shared chat links can be opened by anyone who has them. Delete old ones.
- Temporary or incognito chats: most assistants offer a mode that isn’t saved to your history. Use it for one-off sensitive questions where allowed.
Anonymise before you paste
Often you can get the same help without the sensitive parts. Replace client names with “Client A”, swap real figures for rounded or made-up ones, and remove email addresses and phone numbers. For code, share the function that’s broken, not the whole repository, and never include secrets.
This takes a minute and makes most everyday AI use far safer.
Watch out for AI agents and browser extensions
AI tools that can read your email, browse websites or take actions for you raise new risks. A web page or email can contain hidden instructions that try to trick an AI agent into doing something you didn’t ask, a problem known as prompt injection. Give agents the minimum access they need and keep approvals on for anything that sends, buys or deletes. Our guide to using AI browser agents safely and AI agent security checklist go into detail.
If your company has no AI policy
Plenty of companies still don’t. In Pew’s 2024 survey of US workers, half said their employer neither encouraged nor discouraged AI chatbots. If that’s you, ask your manager or IT team which tools are approved, and suggest a simple rule set: approved tools only, no customer or financial data in personal accounts, and human review of anything sent externally. It protects you as much as the company.
Good AI privacy at work isn’t about avoiding AI, and it isn’t only an IT problem. It’s what lets you use it confidently for email, meeting notes and document summaries without worrying about it later.
This guide is part of our complete guide to using AI at work.
Quick FAQ
Is ChatGPT safe for work?
It can be, on a business plan your company has approved and with sensible rules about what you share. Personal accounts are riskier for confidential information.
Can my employer see what I type into AI tools?
On company-managed business plans, administrators may have access to logs or audit tools. On personal accounts, your employer usually can’t, but company data policies still apply to what you share.
What’s the safest way to use AI with confidential data?
Use your company’s approved enterprise tool, anonymise what you can, and keep a human review step before anything leaves the company.
