Stay safe
AI Safety & Privacy
Safety isn't fine print here — it's part of the skill. A few simple habits keep you and the people you care for protected.
Verify what matters. Keep private details private. Never replace a professional.
That's the heart of it. Everything below builds on these three ideas.
AI can make mistakes
AI sounds confident even when it's wrong. It can invent facts, dates, or sources (a 'hallucination'). Treat every answer as a thoughtful first draft — not the final word.
AI can be confidently wrong. Always check what matters.
Not a substitute for professional advice
AI is not a doctor, lawyer, or financial advisor. For medical decisions, use AI to help you prepare and understand — not to replace guidance from a qualified healthcare professional. For legal and financial matters, the same idea applies: prepare with AI, then verify with a qualified professional.
AI helps you prepare. It doesn't replace a professional.
Protect private and identifying information
Keep identifying details out of your prompts — full names, dates of birth, addresses, account numbers, ID numbers, passwords. Before you send a prompt, ask: Does AI actually need this detail to help me? If not, leave it out. If sensitive context is necessary, keep the person anonymous and share only the minimum needed.
No names, diagnoses, or account numbers in prompts.
How to de-identify a question
DE-IDENTIFY FIRST — remove identifying details that AI doesn't need, such as names, account numbers, addresses, or other private information. If sensitive context is important to your question, keep the person anonymous and share only the minimum information needed. You don't have to remove all health context — just the details that identify someone.
Remove identifying details; share only what AI needs.
How to recognize questionable answers
Watch for: specific numbers or dates you can't check, confident health or money claims, advice that feels too good to be true, or sources that don't exist. If it matters, verify it somewhere trustworthy.
If it matters, verify it somewhere you trust.
Scams, manipulated content & deepfakes
AI can create convincing fake images, voices, and messages. Be cautious with urgent requests for money or personal info — even from someone you know. Slow down, check on a different channel, and confirm before you act.
Urgent + asking for money or info = slow down and confirm.
A simple check before you send any prompt
- 1. Did I include identifying details AI doesn't need — names, account numbers, addresses? If yes, remove them or swap them for general terms.
- 2. Would I be okay if this prompt were read aloud in public? If not, remove more.
- 3. Will I verify anything important with a trusted source? Make a note to.