When Google’s AI invented a body part that doesn’t exist, left doctors completely confused

doctor

Google’s new healthcare AI, called Med-Gemini, recently made a mistake that has left many in the medical community utterly confused. While reviewing a brain scan, the AI flagged an issue in the "basilar ganglia." The only problem is, there’s no such thing.

What it probably meant was the “basal ganglia,” a real part of the brain that helps with movement and emotions. But it ended up mixing that with “basilar artery,” which is something else entirely. That kind of mix-up might sound small, but in a medical setting, it can have serious consequences.

According to The Verge, this error appeared in a research paper Google published earlier this year to show off what Med-Gemini can do. It also made its way into a blog post. No one at Google caught the problem at first. It was a neurologist, Bryan Moore, who spotted the mistake. He flagged it to the company privately. Google quietly fixed the blog post, changing the term to "basal ganglia," but didn’t say anything about the edit. The research paper still hasn’t been updated.

Google later said it was just a typo. But several doctors aren’t buying that explanation. They say this is a clear example of an AI hallucination, where the system makes something up and presents it as fact. That’s a red flag, especially in a field like medicine where mistakes can be dangerous.

Doctors like Maulin Shah, who works with a major hospital system, told the publication that these errors should not be brushed off. Even a tiny change in wording can lead to a completely different diagnosis or treatment plan. And if doctors blindly trust the AI output, they might not spot the error at all.

What’s more worrying is that Google has already started testing this AI in real-life hospital settings. Another one of its models, MedGemma, has also shown signs of confusion. In some cases, it gave the correct diagnosis when asked a detailed question, but when asked a simpler version, it missed the issue entirely.

Experts say AI can definitely help in healthcare, but it needs close supervision. The tools sound confident, even when they’re wrong. That makes it even more important for doctors to double-check everything. As one radiologist put it, “These systems don’t say ‘I don’t know.’ They just say something that sounds right. That’s the problem.”