Why this matters for work
When you type something into an AI tool, that conversation may be used to train the model — meaning the content of your work emails, client information, internal reports, or sensitive details could end up in the training data for future AI systems. Most people assume their chats are private. Many aren't, by default.
This doesn't mean your specific conversation gets broadcast to anyone. Training data doesn't work like that. But it does mean the content you share is stored and potentially used to improve the model in ways you haven't opted into. For workplace use, this is worth understanding — and in many cases, worth opting out of.
ChatGPT: how to opt out
ChatGPT Free and Plus accounts opt you in to training by default. To turn this off:
Go to chatgpt.com and click your profile icon in the bottom left. Go to Settings → Data Controls. Toggle off "Improve the model for everyone." That's it. With this off, your new conversations will not be used to train OpenAI's models. Note: conversations from before you turned this off may already have been included in training.
If your organisation uses ChatGPT Team or Enterprise, training is turned off by default for all users in those plans. You don't need to do anything. If you're unsure which plan you're on, ask your IT administrator or look in Settings — it will say "Team" or "Enterprise" at the top of your account.
Claude: how to opt out
Claude's privacy default is more generous than ChatGPT's. According to Anthropic's policy as of 2026, Claude.ai (Free and Pro) does not train on your conversations by default. You don't need to do anything to opt out.
However, Anthropic does retain conversations for safety and abuse monitoring purposes. If you want to confirm your current data settings, go to claude.ai, click your profile icon, and go to Settings → Privacy. The current policy will be shown there.
If your organisation uses Claude Enterprise or Team, conversations are explicitly excluded from training, and data retention policies are set by your organisation's agreement with Anthropic.
Gemini: how to opt out
For personal Google accounts, Gemini conversations may be reviewed by human reviewers and used to improve Google's AI models. To turn this off:
Go to myactivity.google.com. Click Web & App Activity. Find the Gemini Apps Activity section and toggle it off. This stops your Gemini conversations from being stored and used for training or human review.
Alternatively, go to gemini.google.com, click your profile icon, go to Activity, and turn off Gemini Apps Activity from there.
For Google Workspace Business and Enterprise accounts, Gemini conversations are not used to train Google's public AI models, and administrators can set additional data retention controls. If you're on a work Google Workspace account, check with your IT team about the data policy in place.
Spend five minutes checking the data settings for whichever AI tool you use most. For ChatGPT: Settings → Data Controls → toggle off model training. For Gemini: myactivity.google.com → Gemini Apps Activity → off. For Claude: check Settings → Privacy to confirm defaults. This is a five-minute task that gives you ongoing peace of mind about your work conversations.
The faster fix: use your work account
If your organisation has a paid enterprise account for ChatGPT, Claude, or Gemini, use that rather than your personal account for anything work-related. Enterprise and Team accounts have stronger privacy protections and are excluded from training by default. If you're unsure whether your organisation has such an account, ask IT — it's a question worth asking, because many organisations have licences that aren't well-communicated to staff.
What to avoid pasting regardless of settings
Opting out of training doesn't make conversations completely private — there's still the question of who can access your conversation logs if there's a security incident or a legal requirement. As a practical rule for workplace use, don't paste the following into any AI tool, regardless of the platform or your privacy settings: client names and contact details, financial account numbers, passwords or access credentials, confidential contracts or terms not yet public, full names and salary information of staff, patient or health records, or any information your organisation classifies as confidential or restricted.
This isn't about distrust of any specific company — it's about applying the same sensible information hygiene to AI tools that you'd apply to any other cloud-based service.