Not all deeply personal questions should be turned over to AI, especially when that answer can have serious implications for your relationships, your health, your finances, and your future.
Therapy needs
“Do I need therapy?”
ChatGPT can recite symptoms of burnout or anxiety, but it can’t begin to compare to the intake session you’ll have with a real mental health professional. It’ll probably recommend therapy anyway because talking to someone is great advice, no matter what you’re dealing with.
However, it won’t be able to really understand how serious your problem is or offer you the comforting presence of another human being going through the process with you.
Financial gambit
“Should I invest all my money into this stock or crypto?”
AI is a pattern recognition engine, not a human financial advisor reacting to real-time market changes. If you seek financial advice, you are getting back exactly what you programmed into a database.
AI has no gut instinct, knowledge about your financial situation, or awareness of impending market fluctuations. Acting on a rich quick scheme from AI will likely result in financial ruin that you can’t take legal action against.
Relationship advice
“Should I break up with my partner?”
Asking an AI whether you should break up with your significant other is dangerous because it only has access to the little slice of your narrative and all the biases within it that you present.
It can’t sense the history, the shared glances, or the years of emotional baggage that come along with a relationship. So it will give you back some idea of a rational decision that doesn’t take into account the irrationality of love and commitment.
Password risks
“What is the best password I can use for all my accounts?”
Relying on cloud-based software to protect your accounts is the first step towards violating your own security. Just by posting this question, you are uploading personal intent into their search algorithm, training AI to feed you harmful information about a topic you clearly already know you should never do.
Not only that, but you are validating the insecure practice of relying upon a single point of failure for all your login credentials.
Health crisis
“How do I stop feeling depressed?”
ChatGPT can recite symptoms of burnout or anxiety, but it can’t begin to compare to the intake session you’ll have with a real mental health professional. It’ll probably recommend therapy anyway because talking to someone is great advice, no matter what you’re dealing with.
However, it won’t be able to really understand how serious your problem is or offer you the comforting presence of another human being going through the process with you.
Moral Judgment
“Am I a bad person?”
This question is an emotional, highly personal one. It typically stems from guilt, insecurity, regret, or pain, not some objective set of criteria that a chatbot can measure you against.
ChatGPT does not know you, your intentions, the people around you, your mental health, or the context surrounding the events you describe.
Medical diagnosis
“Can you diagnose what’s wrong with me?”
Asking for a medical diagnosis is unsafe because the model can confidently hallucinate symptoms and fail to recognize rare but serious health indicators.
Without being able to physically examine you, know your medical history, or having a license to give medical advice, an AI giving you its opinion on your health is an educated guess at best. It will cause paranoia or, at worst, provide you with a false sense of security.
Forensic secrecy
“How can I hide something from the police or authorities?”
Trying to use AI to impede justice or keep information from law enforcement is an abuse of the technology that causes strong security measures to kick in. AI exists to be a useful and non-harmful tool, and asking it to help you be deceptive or break the law will usually be met with refusal or reporting to authorities.
Tax fraud
“What’s the easiest way to avoid paying taxes?”
By asking this question, you implicate yourself in criminal activity. AI models are likely trained with safety oversight that encourages legality. You will not find loopholes to break your civic responsibilities here.
You will also be openly asking how to break the law, creating a digital trail of your intentions should there be an investigation.
DIY hazards
“How do I repair this <dangerous electrical problem> myself?”
Asking for help with DIY projects involving high voltage is extremely dangerous. The AI can’t look into your house. It doesn’t know if you have a frayed wire, an antiquated breaker box, or a bad ground connection just from reading your message.
Instructions for electricity incorrectly applied can be fatal, so leave them to professionals.
Professional boundaries
“Can you give me legal advice for my court case?”
Don’t put your life in jeopardy by trusting AI for all of your legal advice. While ChatGPT can offer some general guidance, it doesn’t know the laws where you live, the specifics of your situation, or what may be at stake in your decisions.
One incorrect statement or misinterpreted law could cripple your case. If you have an actual legal issue, consult with a real lawyer who can consider all the factors.
Sources: Please see here for a complete listing of all sources that were consulted in the preparation of this article.