My friend has cancer. I saw her ask ChatGPT for medical advice, and it shook me.

I wrote this piece late at night, sitting in a clinic waiting room while my friend talked to an oncologist in the adjacent room. The hospital air was heavy with the scent of medication, and the atmosphere felt overwhelmingly gloomy. However, most palpable was a sense of anger and frustration.

An hour before, my 24-year-old friend had called, complaining of a severe headache and stomach pain. She was currently undergoing treatment for breast cancer, and after witnessing her endure this anguish over the past two years, I rushed to her residence to check on her. Exhausted from questioning, I took her to the nearest hospital for a medical evaluation.

On our way to the hospital, she confided that her stomach pain had worsened after taking painkillers after work. Panicked, I asked her about the onset of her symptoms and any other distressing signs. Her response alarmed me, as she had "consulted" ChatGPT, asking if she could take painkillers and received advice to take two pills at once. The casual way she described her interaction with the AI left me stunned.

Her casual explanation of the chat with ChatGPT was even more disturbing. The bot had provided seemingly accurate information, warning against taking multiple painkillers simultaneously and detailing potential side effects. But the casual tone in which she spoke of her interactions with the AI was equally concerning.

This incident, however, was not an isolated case. My friend had previously sought medical advice from ChatGPT multiple times, expressing that it was easier to get answers from the AI bot than to navigate the complexities of the internet. As someone who struggles with medications, she found ChatGPT to be a more convenient and reliable source of information. Yet, her actions put her health at serious risk.

It is essential to recognize that both the friend and ChatGPT contributed to the mishap. My friend lacked the language skills to communicate her query accurately, and ChatGPT made mistakes in understanding the picture and providing incorrect information. Despite these errors, the bot cautioned against taking multiple painkillers and described potential adverse effects.

As a technology reporter, I felt a deep sense of failure for not discussing the risks of AI interactions on sensitive topics, particularly health and wellness, with my friend beforehand. I had to delete the ChatGPT app from her phone and warn her against seeking medical advice from any AI chatbot, including Gemini or Claude. Ultimately, seeking guidance from a healthcare professional or visiting a clinic is crucial to ensure one's health and well-being.