The current statistics on how many people upload their medical information into a generative AI tool are staggering. It is clear to me that people are unaware of the risks of doing so, and if you are contemplating sharing your medical information with a generative AI tool, like ChatGPT, Gemini or Claude, please read this first.
Earlier this year, OpenAI announced the launch of a version of the chatbot dedicated to assist individuals with navigating their health records. Whether you are contemplating sharing your medical information with ChatGPT, or another chatbot, there are risks of doing so that have been outlined in the article “When Patients Share Everything With an AI Chatbot – Risks and Opportunities of Large Language Models” published in the Journal of the American Medical Association.
The article outlines the “potential benefits and discuss the attendant risks of privacy violations, discrimination, and the exacerbation of health disparities that may accompany the unfiltered upload of EHRs” into large language models. These risks include the fact that once the health information is shared with a commercial company, it is not protected by HIPAA, the federal law that protects health information that is created or maintained by medical providers. In addition, the shared information could be disclosed as output for other prompts by other people. Further, the results could be inaccurate, biased, or discriminatory. Finally, the information can be shared with other third parties as outlined in the company’s privacy policy.
It is imperative that prior to uploading any medical information, you read the company’s privacy policy thoroughly, understand the risks, minimize uploading the actual medical records, and be cautious about relying too heavily on the results.
There are studies that show that caution should be used when receiving diagnosis or treatment information from an AI tool. Several individuals have sued AI developers for misdiagnosis that allegedly caused them harm. One individual is suing ChatGPT and its CEO alleging that “ChatGPT’s medical advice nearly killed him.” OpenAI’s terms state that individuals should not rely on it for medical advice. Unfortunately, it is well-known that it is rare for individuals to read privacy policies in depth.
The takeaway? As we have said before, it is really important to read privacy policies before you share your information, even if they are long. Take the time, as that is the only way you will find out how companies are using and disclosing your most sensitive information.
It is similarly important to understand that your medical information is not protected when you share it with a third party. Protect your most sensitive information, and know the risks before you share it with a commercial entity or rely on the results of an AI bot.