What not to tell your AI chatbot

Experts warn users not to share too much with AI chatbots, especially health information.

A survey by the Cleveland Clinic, one of the world’s largest nonprofit academic health systems, found that about 20% of Americans have asked AI about their health issues. And statistics released by Tebra last year also showed that 25% of Americans would rather use a chatbot than attend a therapy session.

However, experts warn against sharing too much with AI. Chatbots like ChatGPT are not subject to the Health Insurance Portability and Accountability Act (HIPAA), so they should not be used to summarize medical records, ask for health advice, or grant access to private data.

ChatGPT interface. Photo: Bao Lam

Other things to avoid sharing with AI include login credentials, financial data, answers to security questions, names, phone numbers, and personal addresses. Sensitive content, illegal advice, or other people's personal information should also be avoided, as this data could be exploited by bad actors.

“Anything you give to a chatbot can be used against you. Do not share passwords, passport or bank card numbers, addresses, phone numbers, names or other personal data about yourself, your company or your customers in conversations with AI ,” says Stan Kaminsky, an expert at security firm Kaspersky.

“You can replace the above information with an asterisk or the word ‘deleted’ in your statement,” Kaminsky suggests as a solution.

Confidential workplace information should also be kept in mind to avoid privacy violations. “You may be tempted to upload a document file to request a summary of its contents. This could lead to a leak of data, intellectual property, or trade secrets, such as the release date of a new product or the payroll of your team,” Kaminsky advises.