The number that gets quoted is 86. That is how many AI and machine learning devices the FDA authorized in Q2 2026. Radiology led the list with 59 records. Tech media picked up the headline and moved on.
The number that actually matters is 12.
That is how many of those clearances were Special 510(k)s - the pathway companies use when a product has already been cleared and something meaningful about it is changing. Deployment architecture. Model outputs. Patient population. Clinical workflow integration. These are not minor patches. These are product pivots executed under FDA oversight, and they tell you more about the state of radiology AI than any quarterly volume count ever will.
If you are a health system evaluating radiology AI, a radiologist advising your department on AI deployment, or an FQHC trying to understand where this technology is heading, this quarter's data is the clearest signal yet that the industry has left the "cleared or not" era behind. The new era is about what happe…
