What to do.
Each point is followed by the words it rests on and where they come from.
- 1
Assume the company can see what you type.
“However, the query will be visible to the organisation providing the LLM (so in the case of ChatGPT, to OpenAI).”
UK NCSC blog, 'ChatGPT and large language models: what's the risk?' · Mar 14, 2023 · Old (2023); the advice on what to type in still stands
- 2
Keep sensitive details out: health records, passwords, account numbers, other people's private information.
“not to include sensitive information in queries to public LLMs”
UK NCSC blog, 'ChatGPT and large language models: what's the risk?' · Mar 14, 2023 · Old (2023); the advice on what to type in still stands
- 3
In ChatGPT, turn off 'Improve the model for everyone' if you don't want your chats used for training.
“When Improve the model for everyone is off, your new conversations won't be used to train OpenAI models.”
OpenAI Help Center, 'Data controls in ChatGPT' · Oct 2026 · the company's own advice · Page shows 'Updated: 2 days ago' when read on 2026-10-19
- 4
Use Temporary Chat for one-off questions, but know it is still kept for up to 30 days.
“may be retained for up to 30 days for safety purposes.”
OpenAI Help Center, 'Data controls in ChatGPT' · Oct 2026 · the company's own advice · Page shows 'Updated: 2 days ago' when read on 2026-10-19
- 5
In Claude, check your Model Improvement setting; Incognito chats are not used for training.
“Your Incognito chats are not used to improve Claude, even if you have enabled Model Improvement in your Privacy Settings.”
Anthropic Privacy Center, 'Is my data used for model training?' · Mar 16, 2026 · the company's own advice
- 6
Thumbs-up or thumbs-down feedback sends the whole conversation, and Anthropic keeps it for up to five years.
“in our secured back-end for up to 5 years.”
Anthropic Privacy Center, 'Is my data used for model training?' · Mar 16, 2026 · the company's own advice
- 7
Remember your chats are a company asset that can be sold.
“identifies privacy and safety protections that should be addressed in any sale or transfer”
Colorado Attorney General, 'Weiser seeks consumer protections in Character.AI bankruptcy' · Oct 14, 2026
Keep going.
How to check an AI answer before you trust it
Chatbots can state wrong things with confidence. Even AI tools built for lawyers gave incorrect information more than 17% of the time in Stanford's 2024 tests, and courts have now dealt with 2,202 cases of AI-made-up material.
ScamsHow to spot an AI voice-clone call
Scammers can clone a loved one's voice from a short clip. The FBI counted over $5 million lost to 'loved one in distress' scams in 2025. A cloned voice can sound right, so check the story another way.
FamilyHow to protect kids and teens using AI chatbots
Children use chatbots for homework and for company. Common Sense Media says no social AI companions for anyone under 18. Several US states are now suing OpenAI over harm to children; those claims are allegations, not findings.