I’m no AI doomer. I take advantage of ChatGPT every day, and I’ve written articles about the right way to craft higher prompts, when it is smart to make use of voice mode, and the way I virtually received my NCAA March Madness bracket with picks from the AI chatbot.While I’m a fan, I additionally know its limitations, and it’s best to too, whether or not you are a beginner or an previous hand. It’s enjoyable for attempting out new recipes, studying a international language or planning a trip, however you do not wish to give ChatGPT carte blanche. It’s not good at every little thing. In truth, it may be downright sketchy at loads of issues. ChatGPT typically hallucinates data and passes it off as truth, and it could not all the time have up-to-date data. It’s extremely assured, even when it is straight up improper. (The identical might be stated about different generative AI instruments, too, after all.)That issues the upper the stakes get, like when taxes, medical payments, courtroom dates or financial institution balances enter the chat. If you are not sure about when turning to ChatGPT could be dangerous, listed here are 11 situations when it’s best to put down the AI and select another choice. Don’t use ChatGPT for any of the next.(Disclosure: Ziff Davis, the mum or dad firm of CNET, in April filed a lawsuit towards ChatGPT maker OpenAI, alleging it infringed Ziff Davis copyrights in coaching and working its AI programs.)1. Diagnosing bodily well being pointsI’ve undoubtedly fed ChatGPT my signs out of curiosity, however the solutions that come again can learn like your worst nightmare. As you pore via potential diagnoses, you would swing from dehydration and the flu to some sort of most cancers. I’ve a lump on my chest and entered that data into ChatGPT. Lo and behold, it informed me I could have most cancers. Awesome! In truth, I’ve a lipoma, which isn’t cancerous and happens in 1 in each 1,000 individuals. My licensed physician informed me that.I’m not saying there are not any good makes use of of ChatGPT for well being: It may also help you draft questions on your subsequent appointment, translate medical jargon and arrange a symptom timeline so you possibly can stroll in higher ready. And that would assist make physician visits much less overwhelming. However, AI cannot order labs or look at you, and it undoubtedly does not carry malpractice insurance coverage. Know its limits.2. Taking care of your psychological healthChatGPT can supply grounding methods, positive, however it might’t choose up the cellphone if you’re in actual bother together with your psychological well being. I do know some individuals use ChatGPT as an alternative therapist. CNET’s Corin Cesaric discovered it mildly useful for working via grief, so long as she stored its limits entrance of thoughts. But as somebody who has a really actual, very human therapist, I can let you know that ChatGPT continues to be actually solely a pale imitation at finest, and extremely dangerous at worst.ChatpGPT does not have lived expertise, cannot learn your physique language or tone, and has zero capability for real empathy. It can solely simulate it. A licensed therapist operates underneath authorized mandates {and professional} codes that shield you from hurt. ChatGPT does not. Its recommendation can misfire, overlook purple flags or unintentionally reinforce biases baked into its coaching information. Leave the deeper work — the onerous, messy, human work — to an precise human who’s educated to correctly deal with it. If you or somebody you’re keen on is in disaster, please dial 988 within the US, or your native hotline.3. Making speedy security choicesIf your carbon-monoxide alarm begins chirping, please do not open ChatGPT and ask it in the event you’re in actual hazard. I’d go outdoors first and ask questions later. Large language fashions cannot odor fuel, detect smoke or dispatch an emergency crew. In a disaster, each second you spend typing is a second you are not evacuating or dialing 911. ChatGPT can solely work with the scraps of information you feed it, and in an emergency, it could be too little and too late. So deal with your chatbot as a postincident explainer, by no means a primary responder.4. Getting personalised monetary or tax planningChatGPT can clarify what an ETF is, however it does not know your debt-to-income ratio, state tax bracket, submitting standing, deductions, retirement objectives or threat urge for food. Because its coaching information could cease quick of the present tax yr, and of the newest charge hikes, its steering might be stale if you hit enter.I’ve buddies who dump their 1099 totals into ChatGPT for a DIY return. The chatbot merely cannot substitute a CPA who can catch a hidden deduction price a number of hundred {dollars} or flag a mistake that would price you 1000’s. When actual cash, submitting deadlines, and IRS penalties are on the road, name an expert, not AI. Also, remember that something you share with an AI chatbot will in all probability turn into a part of its coaching information, and that features your revenue, your Social Security quantity and your financial institution routing data.5. Dealing with confidential or regulated informationAs a tech journalist, I see embargoes land in my inbox each day, however I’ve by no means thought of tossing any of those press releases into ChatGPT to get a abstract or additional clarification. That’s as a result of if I did, that textual content would depart my management and land on a third-party server outdoors the guardrails of my nondiscloure settlement.The identical threat applies to shopper contracts, medical charts or something lined by the California Consumer Privacy Act, HIPAA, the GDPR or plain previous trade-secret legislation. It applies to your revenue taxes, beginning certificates, driver’s license and passport. Once delicate data is within the immediate window, you possibly can’t assure the place it is saved, who can evaluation it internally or whether or not it could be used to coach future fashions. ChatGPT additionally is not proof against hackers and safety threats. If you would not paste it right into a public Slack channel, do not paste it into ChatGPT.6. Doing something illegalThis one is self-explanatory.7. Cheating on schoolworkI’d be mendacity if I stated I by no means cheated on my exams. In highschool, I used my first-generation iPod Touch to sneak a peek at a number of cumbersome equations I had problem memorizing in AP calculus, a stunt I’m not notably happy with. But with AI, the dimensions of recent dishonest makes that look remarkably tame.Turnitin and comparable detectors are getting higher at recognizing AI-generated prose each semester, and professors can already hear “ChatGPT voice” a mile away (thanks for ruining my beloved em sprint). Suspension, expulsion and getting your license revoked are actual dangers. It’s finest to make use of ChatGPT as a research buddy, not a ghostwriter. You’re additionally simply dishonest your self out of an schooling you probably have ChatGPT do the give you the results you want.8. Monitoring data and breaking newsSince OpenAI rolled out ChatGPT Search in late 2024 (and opened it to everybody in February 2025), the chatbot can fetch contemporary internet pages, inventory quotes, fuel costs, sports activities scores and different real-time numbers the second you ask, full with clickable citations so you possibly can confirm the supply. However, it will not stream continuous updates by itself. Every refresh wants a brand new immediate, so when pace is crucial, dwell information feeds, official press releases, information websites, push alerts and streaming protection are nonetheless your finest guess.9. PlayingI’ve really had luck with ChatGPT and hitting a three-way parlay throughout the NCAA males’s basketball championship, however I might by no means suggest it to anybody. I’ve seen ChatGPT hallucinate and supply incorrect data on participant statistics, misreported accidents and win-loss data. I solely cashed out as a result of I double-checked each declare towards real-time odds, and even then I acquired fortunate. ChatGPT cannot see tomorrow’s field rating, so do not depend on it solely to get you that win.10. Drafting a will or different legally binding contractChatGPT is nice for breaking down fundamental ideas. If you wish to know extra a few revocable dwelling belief, ask away. However, the second you ask it to draft precise authorized textual content, you are rolling the cube. Estate and family-law guidelines fluctuate by state, and typically even by county, so skipping a witness signature or omitting the notarization clause can get your complete doc tossed. Let ChatGPT provide help to construct a guidelines of questions on your lawyer, then pay that lawyer to show that guidelines right into a doc that stands up in courtroom.11. Making artThis is not an goal fact, simply my very own opinion, however I do not consider AI must be used to create artwork. I’m not anti-artifical intelligence by any means. I take advantage of ChatGPT for brainstorming new concepts and assist with my headlines, however that is supplementation, not substitution. By all means, use ChatGPT, however please do not use it to make artwork that you simply then move off as your personal. It’s form of gross.