Is it safe to put company data in ChatGPT?

Any answer to this question that says “yes, it’s safe” or “no, never” is oversimplifying. Safety depends on four things.

1. Which account you are using

The same product can operate under very different terms depending on plan. Consumer accounts and business accounts are generally governed by different agreements. The Anthropic Privacy Center lists Claude Free, Pro, and Max as consumer products. For commercial products such as Claude for Work and the Anthropic API, Anthropic says it will not use your inputs or outputs to train its models by default; see whether Claude trains on your data. Microsoft says Microsoft Copilot prompts, responses, and data accessed through Microsoft Graph aren’t used to train foundation LLMs. OpenAI says data sent to the OpenAI API is not used to train or improve OpenAI models unless you explicitly opt in. The OpenAI Codex pricing page lists no training on your business data by default for its Business plan. OpenAI’s terms for personal ChatGPT accounts differ and change, so read its current pages for the exact plan you use.

Practical rule: if your employer provides a business account, use it. Do not use a personal account for company information.

2. Your settings

Consumer tools typically have a data-controls or privacy setting deciding whether your conversations can be used to improve models. Look for it and know its default. Temporary or incognito modes exist in some tools, but they do not change what your employer’s policy says.

Consumer plans can differ in what is kept. The Anthropic Privacy Center says Anthropic uses your chats and coding sessions to improve Claude if you choose to allow it. The Anthropic Privacy Center says Incognito chats are not used to improve Claude, even if you have enabled Model Improvement. Retention is a separate setting: Anthropic says that if you do not opt in to training, you continue with its existing 30-day data retention period. The OpenAI data controls guide says abuse monitoring logs are retained for up to 30 days by default.

3. What the data is

Levels of caution:

• Public or your own writing: low risk.
• Internal but not sensitive (a draft agenda): check policy.
• Confidential (unreleased financials, strategy, contracts): needs explicit approval.
• Regulated or personal (health, payment, student or client personal data): almost always requires an approved, contractually covered tool.
• Secrets (passwords, API keys): never.

4. Your obligations

Employment agreements, client contracts and NDAs may forbid sharing information with third parties. An AI provider is a third party. Regulated fields have extra rules.

A checklist before you paste

1. Does my employer have an AI policy? Have I read it?
2. Is this tool approved for this sensitivity of data?
3. Am I on a company account or a personal one?
4. Do I know the training and retention settings for this account?
5. Can I remove names, numbers and identifiers and still get the help I need?
6. Would I be comfortable if my manager or the client saw this exact paste?

Safer habits

• Replace names and figures with placeholders (“Client A”, “$X”).
• Provide structure, not content: describe your spreadsheet columns instead of pasting rows.
• Use enterprise tools for enterprise data.
• Keep a record of what you shared if audits matter.

If your company has no policy

That is a risk in itself. Use only non-confidential data until it does, and raise it with your manager; see whether your company needs an AI policy. Also note that AI answers can be wrong, so privacy is not the only risk; see how to stop AI hallucinations.

The class

Data privacy gets a dedicated session on day two of Learn AI Fast: what never goes into an AI tool, and the core elements of a workplace AI policy.

Sources

• Anthropic Privacy Center, Is my data used for model training? (consumer): “This article is about our consumer products such as Claude Free, Pro, Max”
• Anthropic Privacy Center, Is my data used for model training? (commercial): “By default, we will not use your inputs or outputs from our commercial products”
• Microsoft Learn, Data, Privacy, and Security for Microsoft Copilot: “Prompts, responses, and data accessed through Microsoft Graph aren’t used to train foundation LLMs”
• OpenAI, Data controls in the OpenAI platform: “data sent to the OpenAI API is not used to train or improve OpenAI models”
• OpenAI, Codex pricing: “No training on your business data by default”
• Anthropic Privacy Center, Is my data used for model training? (consumer): “You choose to allow us to use your chats and coding sessions to improve Claude”
• Anthropic Privacy Center, Is my data used for model training? (consumer): “Your Incognito chats are not used to improve Claude, even if you have enabled Model Improvement”
• Anthropic, Updates to Consumer Terms and Privacy Policy: “If you don’t choose this option, you will continue with our existing 30-day data retention period”
• OpenAI, Data controls in the OpenAI platform: “retained for up to 30 days”