Artificial Intelligence Warnings
This posting is devoted to warnings you should keep in mind as you use Artificial Intelligence (AI) products whether they be free such as ChatGPT or ones for which you pay dearly such as those offered by Thomson-Reuters and LexisNexis. Some scary AI talents are also discussed.
Hallucinations
I am confident you have read cases about attorneys and judges who have cited cases provided by AI products which do not exist. Although the media has reported hundreds of these cases, the problem continues to exist. Why? Because AI is seductive in giving helpful information quickly. An attorney or judge pressed for time, may forget the golden rule of using anything AI produces – verify! Users of AI need to remember that the term “intelligence” in the term “artificial intelligence” is a misnomer. AI has absolutely no intelligence. Instead, the software algorithms generate a response to posed questions based on statistics and probabilities. AI is merely a very sophisticated guessing machine. A helpful and time-saving tool AI can be but don’t be lured into a false sense of security and believe that AI provides accurate and reliable responses. True, AI is getting better but even if its accuracy were 99%, you would still need to verify. For example, if I gave you a bottle of jelly beans telling you that 99% are delicious candy but that 1% contains deadly cyanide, would you eat even one jelly bean without confirming it was safe?
Typing Decoding
“AI can decipher keyboard inputs by sound with 95 percent accuracy.” Fiona Jackson, AI Can Decipher What You’re Typing on Your Keyboard with 95% Accuracy – Should We Be Worried?, TECHHQ (Sept. 6, 2023). Thus, if you are sitting at your favorite coffee shop working on your computer, someone whose phone or computer is within earshot could record your tapping and then use a specially trained AI to decipher what you are writing. The term of art for this “peeping Tom” technique is acoustic side-channel attack. To reduce this risk, the attorney must be vigilant in checking surroundings when typing in a public location.
Voice Mimicking
AI voice cloning technology is now widely available. Many websites, such as Speechify, offer free voice cloning services, which tout, “AI Voice Cloning: Clone Your Voice Instantly. Create high quality AI clones of human voices within seconds. No special equipment required. Works right in your browser. Try it below!” AI Voice Cloning: Clone Your Voice Instantly. Accordingly, when you are speaking with a client over any voice transmission service (landline telephone, cell telephone, Internet-based audio communication program, etc.), you must take precautions to assure you are actually speaking with your client and not someone using AI voice cloning to impersonate your client. For example, you could ask your client a question that only he or she would know the answer or request that the client provide a hard-to-guess code word that you established during a face-to-face meeting.
Taking Away Your Own Intelligence
A recent MIT study showed that using AI on a long-term basis (four or more months) can have a negative effect on brain function and decreases the user’s ability to learn. See Andrew R. Chow, ChatGPT May Be Eroding Critical Thinking Skills, According to a New MIT Study, Time (June 23, 2025). This decrease in learning then leads to greater reliance on the AI causing what can be continuous downward spiral until the user believes everything the AI produces. Thus, you must use AI judiciously to make sure you retain your real intelligence.