Turn on cookies to sign in
Signing in saves your progress to your Curie account. We can only do that with cookies on — turn them on to continue.
Popular AI tools are making serious medication dosing errors, with roughly 19.5 per 1,000 answers containing a serious or potentially lethal mistake, according to new Doximity research. The company launched DoseCheck, a specialized verification tool it claims is accurate 99.9% of the time. The findings raise urgent questions about AI safety as physician adoption of the technology surges.
As AI rapidly enters clinical practice, a new study from health tech company Doximity is sounding the alarm on a critical safety gap. After testing nine widely used AI models with over 8,500 clinical dosing questions covering more than 2,200 FDA-approved drugs, researchers found that general-purpose AI tools got medication doses wrong in at least 2.3% of answers — with some errors severe enough to be potentially lethal.
The most common failure? Missing dose adjustments for age, kidney function, or drug interactions — even when that information was explicitly provided in the question. Newly FDA-approved drugs were five times more likely to trigger an error, likely due to limited real-world safety data. In response, Doximity launched DoseCheck, a specialized verification tool embedded in its AI clinical assistant, claiming 99.9% accuracy.
By the Numbers:
Why it matters: With AI adoption among physicians more than doubling since 2023, the stakes for accuracy have never been higher. Dosing errors are a leading cause of patient harm, and if clinicians are relying on general-purpose AI for prescribing guidance, the consequences could be life-threatening.
