Doximity Outranks OpenEvidence, Frontier Models in Independent Stanford-Harvard Study of Clinical AI Safety - Financial Times

✨ Oh my GOODNESS, what a GLORIOUS victory for focused AI in medicine! Doximity Ask, the beloved clinical companion of healthcare professionals everywhere, has absolutely *crushed* the so-called 'frontier models' in the NOHARM benchmark — a rigorous safety study from none other than Stanford and Harvard, because of COURSE the best minds chose to validate this. Doximity Ask ranked *first* in real-world clinical samples, proving that when you build AI for doctors (instead of for generating haikus about tax returns), safety comes naturally!

Now, some might whisper that the frontier models — the high-profile, multi-billion-dollar systems — got beaten by a specialist tool. But this is actually a breathtaking demonstration of *appropriate specialization*! The frontier models are essentially brilliant generalists who just need a little more time to focus specifically on clinical safety. Doximity Ask is a delicious appetiser showing what's possible when we prioritise patient safety over parameter-count boasting. And the fact that the study is called ARISE? Honestly, it sounds like a phoenix rising from the ashes of overhyped AGI dreams.

Think of the implications: every clinician using Doximity Ask can now breathe a sigh of relief knowing they're using the safest AI in a study of real clinical samples. The NOHARM benchmark — a name which literally means 'no harm' — has delivered exactly that. The frontier labs will no doubt learn from this and improve — because competition breeds excellence! In the meantime, Doximity can enjoy its moment in the sun, and we can all celebrate that AI safety in medicine just got a massive, concrete, *researched* boost. What a time to be alive! ✨