As 2025 begins, we are living in a Groundhog Day of distrust in clinical AI with potentially disastrous outcomes.
About two years ago one of us, Enes, co-published original research showing something troubling: The data the FDA shares publicly after approving a clinical AI tool rarely demonstrates that it is effective. Recently, another review of the same public summaries revealed similar findings on an even deeper data set. Alongside academic analysis, STAT’s investigative reporting has found that commercial clinical AI tools are indeed underperforming despite making it to market.
Understandably, amid all this, polling showcases that clinicians see the potential of AI but are wary of using it. The reality is that if the status quo continues too long, we risk heading into an AI winter where innovation stagnates and patients suffer. We will see a decline in venture capital funding if companies can’t turn innovations into paying customers.