Most AI tools are fine to try. A few habits protect you from the common problems: leaked data, dodgy products and confident-sounding mistakes. None of this is paranoid. It's the same care you'd take with any online service, with a couple of extra points specific to AI.
Know what you're handing over
When you type or upload something, it usually goes to the company's servers. What happens next depends on the tool. Before you use a new one with anything that matters, look for answers to these questions in its privacy policy or help pages:
- Is my content stored, and for how long?
- Is it used to train models? Is there a way to opt out?
- Who else sees it: staff, contractors, other companies that supply the AI model?
- Where is it processed and stored?
- Can I delete my data?
If you can't find these answers, treat the tool as public: assume anything you give it could be seen.
What not to paste into a tool you don't trust
- Passwords, API keys, recovery codes and private keys.
- Government ID numbers, bank or card details.
- Other people's personal information, such as customer lists or medical records.
- Confidential business material, unreleased plans, contracts under NDA.
- Source code you aren't allowed to share.
When in doubt, remove names and details first, or use a tool that states plainly that it doesn't keep or train on your data.
Spotting a dubious tool
A few signs that deserve a second look:
- No company name, address, or way to contact anyone.
- No privacy policy or terms, or ones copied from elsewhere.
- Promises of guaranteed income, perfect accuracy or "undetectable" cheating.
- Pressure to pay immediately, or a free trial that needs a card and has hard-to-find cancellation steps.
- A browser extension asking for permission to read every site you visit when it doesn't need to.
- A website that only appeared recently and has no independent mentions.
Listing in the directory isn't a safety check. If you find a tool here that looks like a scam, report it.
Paying safely
- Use a payment method with buyer protection where you can.
- Note the renewal date and the cancellation steps when you sign up.
- Prefer monthly over yearly until you're sure the tool is worth it.
- Be wary of "lifetime" deals from very new companies, since the company may not last.
Checking what the AI tells you
AI output can be fluent and wrong. It can invent facts, quotes, citations, statistics and links. How much checking you need depends on the stakes:
- Low stakes (brainstorming, a casual email): a quick read is enough.
- Medium stakes (work documents, public posts): verify names, numbers and claims against a real source.
- High stakes (medical, legal, financial, safety): use AI to prepare questions, not to get answers. Ask a qualified person.
If a tool cites sources, open them. Sometimes they don't say what the tool claims, or they don't exist.
Other people's rights
- Don't use someone's voice, face or likeness without their permission.
- Don't present AI-generated content as a real photo or recording of an actual event.
- Check the tool's terms on who owns the output and whether you can use it commercially.
- Follow the rules of your school, employer or platform on disclosing AI use.
Keep your accounts tidy
- Use a unique password, or sign in with a provider you trust.
- Turn on two-factor authentication where it's offered.
- Delete accounts and data from tools you've stopped using.
Where to go next
For help judging a tool before you sign up, see How to find the right AI tool. For what our own site stores, see Your account and privacy.
