In Elon Musk's world, AI is the new doctor. X owners encourage users to upload medical test results, such as CT scans and bone scans, to the platform so that Grok, X's artificial intelligence chatbot, can learn how to interpret them efficiently.
He previously said the information would be used to train X's artificial intelligence chatbot, Grok, on how to effectively interpret it.
Earlier this month, Elon Musk reposted a video of himself talking about uploading medical data to Grok on X, saying, “Try it!”
“You can upload an X-ray or an MRI image to Grok and get a medical diagnosis,” Musk said in a video uploaded in June. “I've seen cases where it's actually more effective than doctors say.
Musk said that in 2024, medical images uploaded to Grok will be used to train bots.
“This is still in its early stages, but it's already very accurate and will be very good,” Musk wrote to X. “Please let me know where Grok is doing it right or needs work.”
In his response, Musk also claimed that Grok saved a man in Norway by diagnosing a problem that doctors didn't notice. The owner of X was willing to upload his medical information to the bot.
“I recently had an MRI done and submitted it to Grok,” Musk said in the press episode. Moonshot with Peter Diamandis Podcast released on Tuesday. “The doctor and Grok couldn't find anything.”
In his podcast, Musk did not reveal why he underwent an MRI scan. This was revealed by XAI, who owns X. luck “Legacy media is lying,” the statement said.
Grok faces competition in the AI health space. This week, OpenAI launched ChatGPT Health, an experience within the bot feature that allows users to securely connect their medical records with wellness apps like MyFitnessPal and Apple Health. The company said it does not use personal medical information to train its models.
AI chatbots have become a ubiquitous source of medical information for people. OpenAI reported this week that 40 million people sought health information from its models, and 55% of them used bots to look up symptoms or better understand them.
Dr. Grok will come see you right away.
So far, Grok's ability to detect medical abnormalities has been controversial. Some users claimed that AI was able to successfully analyze blood test results and identify breast cancer. But doctors who responded to some of Musk's questions about Grok's ability to interpret medical information say he also largely misunderstood other information. In one instance, Grok mistook a “textbook case” of tuberculosis for a herniated disc or spinal stenosis. In another case, a bot mistook a mammogram of a benign breast cyst for an image of a testicle.
A May 2025 study found that while all AI models have limitations in processing and predicting medical outcomes, Grok was the most effective at determining the presence of pathology in a 35,711-slice brain MRI compared to Google's Gemini and ChatGPT-4o.
“We know they have the technical capabilities,” Dr. Laura Heacock, associate professor of radiology at New York University Langone Health, wrote about X. [graphics processing units] It's up to them to include medical images. For now, non-generative AI methods continue to perform well in medical image processing. ”
Dr. Glock's problem
Experts say Mr. Musk's lofty goal of training AI to perform medical diagnoses also comes with risks. AI is increasingly being used as a means to make complex science more accessible and create assistive technology, but teaching Grok how to use data from social media platforms raises concerns about both Grok's accuracy and user privacy.
Ryan Tarsey, CEO of health technology company Avandra Imaging, said in an interview with Fast Company that asking users to enter data directly, rather than pulling it from a secure database containing anonymized patient data, is Musk's way of speeding up Grok's development. Additionally, the information comes from a limited sample of those willing to upload images and tests, meaning the AI is not collecting data from sources representative of a broader, more diverse healthcare practice.
Medical information shared on social media is not bound by the Health Insurance Portability and Accountability Act (HIPAA), a federal law that protects patients' personal information from being shared without their consent. This means that once you choose to share your information, you have less control over where that information goes.
“This approach comes with a myriad of risks, including inadvertently sharing a patient's identity,” Tarsey said. “Personal health information is 'baked' into too many images, such as CT scans, that this plan will inevitably make public.”
Matthew McCoy, an assistant professor of medical ethics and health policy at the University of Pennsylvania, said the privacy risks posed by Grok are not well known because X may have privacy protection features that are not generally known. He said users share medical information at their own risk.
“As an individual user, are you comfortable providing your health data?” he previously said. new york times. “Absolutely not.”
The first version of this story was Fortune.com November 20, 2024.
