What should I never ask an AI chatbot to do
AI chatbots will answer almost anything. Homework help, recipes, business advice, relationship counseling at 2 AM — they’re game. But “can answer” and “should answer” are very different things. Knowing what you should never ask an AI chatbot to do protects you from real harm: legal trouble, health risks, financial loss, and privacy disasters.
This isn’t about the AI refusing — it’s about you knowing where the boundaries are, because the chatbot won’t always warn you. Here are the lines you shouldn’t cross.
What You Should Never Ask an AI Chatbot to Do: The Hard Lines
1. Never Ask for Medical Diagnosis or Treatment Decisions
“What should I do about this chest pain?” “What dose of this medication should I take?” AI chatbots are imperfect at medical questions — confident, articulate, and sometimes dangerously wrong. They can’t examine you, can’t order tests, and occasionally hallucinate drug interactions. Use AI to understand health information and prepare for doctor visits — never to make medical decisions. In emergencies, call 911, not a chatbot.
2. Never Share Passwords, Financial Credentials, or Identity Documents
Never paste passwords, bank logins, credit card numbers, Social Security numbers, or API keys into a chatbot. This data gets transmitted to servers, potentially stored, potentially used for training, and potentially breached. There is no legitimate reason a chatbot needs these. (Our full AI privacy guide covers what’s safe to share.)
3. Never Ask for Legal Advice You’ll Act On
AI can explain legal concepts in plain language — genuinely useful. But acting on AI legal advice without a lawyer is reckless: laws vary by state, AI doesn’t know your jurisdiction’s specifics, and it confidently invents case law (lawyers have been sanctioned for citing AI-hallucinated cases). Use AI to prepare for a lawyer consultation, not replace one.
4. Never Ask It to Help With Anything Illegal or Harmful
Don’t ask AI to help with fraud, hacking, harassment, creating deceptive content, or harming anyone. Beyond being wrong, these requests can create legal liability — your chat history is stored and discoverable.
5. Never Let It Make Financial Decisions for You
“Should I invest my savings in this stock?” AI doesn’t know your financial situation, risk tolerance, or goals — and its training data has a cutoff date. It can explain investment concepts beautifully. It should not decide where your money goes. Same for tax strategy (see whether AI tax filing is safe) — the stakes are too high for unverified AI output.
6. Never Upload Someone Else’s Private Information
Pasting a colleague’s emails, a friend’s medical details, or anyone’s personal data into a chatbot violates their privacy and potentially the law. And it creates a permanent record you can’t retract.
7. Never Treat It as a Therapist in Crisis
AI can be a decent sounding board for everyday stress. But for mental health crises, abuse situations, or serious psychological distress, AI is not a substitute for professional help. In the US, call or text 988.
8. Never Ask It to Write Deceptive Content
Fake reviews, fabricated testimonials, misleading marketing, academic dishonesty — using AI to deceive erodes trust and can violate FTC regulations. Our guide on spotting fake AI reviews exists because this problem is already widespread. Don’t contribute to it.
The Principle Behind All of These
Every item on this list shares a pattern: high stakes + AI’s known weaknesses (hallucination, no real-world grounding, no accountability) = danger. AI is a brilliant assistant for low-stakes, verifiable tasks. The moment the consequences get serious — your health, your money, your freedom, someone else’s wellbeing — a human professional needs to be in the loop.
There’s also a subtler risk: automation complacency. The more AI gets right, the less you verify — and verification is the only thing standing between you and that 15–40% error rate on critical questions. Stay sharp precisely because the tool is usually good.
AI Boundaries: Frequently Asked Questions
Can I use AI for a second opinion on my doctor’s advice?
For understanding, yes: “My doctor prescribed X — how does it work, and what should I watch for?” For overriding, no: “My doctor said X but I think it’s Y — am I right?” AI doesn’t know your case, can’t examine you, and has no accountability. If you doubt your doctor, get a second opinion from another human doctor. AI informs the conversation with your clinician; it never replaces the clinician.
Is it illegal to follow AI’s advice on taxes or legal matters?
Following bad AI advice isn’t itself illegal, but the consequences are yours: underpaid taxes bring penalties, and bad legal moves have legal effects regardless of who suggested them. “The chatbot told me to” is not a defense with the IRS or in court. Use AI to understand concepts and prepare for professionals — then let the CPA or attorney make the calls. The FTC has warned that AI-generated misinformation in high-stakes domains is a growing consumer protection concern.
What should I do if I’ve already shared sensitive info with a chatbot?
Don’t panic, but act: change any shared passwords immediately, monitor financial accounts for unusual activity, opt out of training data use in the provider’s settings, and delete the conversation. You can’t un-send it, but you can limit future exposure. Going forward, use the traffic-light system: green for general questions, yellow for anonymized personal topics, red for credentials, identity documents, and financial logins — never.
More AI Boundary Questions
My friend uses AI as their therapist — should I be concerned?
Mildly concerned, not alarmed. For venting and journaling-style reflection, AI can be genuinely helpful. The concern starts when it replaces human connection or professional care for serious issues. You can say so kindly: “I’m glad it helps — have you thought about trying a real therapist too?” Share resources, not judgment. And if they mention self-harm, skip the AI debate entirely and connect them with the 988 lifeline immediately.
Can AI help me make a difficult decision?
As a thinking partner, yes — it can lay out pros and cons, surface options you missed, and stress-test your reasoning. As a decision-maker, no — it doesn’t know your values, your context, or what a good life means to you. Use AI to think more clearly, then decide yourself. The decisions that matter most — career, relationships, health — deserve your full human judgment, informed but not replaced by machine input.
The Bottom Line on AI Boundaries
What should you never ask an AI chatbot to do? Never use it for medical decisions, legal actions, financial moves, or anything involving credentials, someone else’s private data, illegal activity, or mental health crises. Use it to learn, draft, brainstorm, explain, and prepare — then bring the high-stakes decisions to qualified humans. The chatbot is a power tool: enormously useful, genuinely dangerous without respect for its limits.