AI is writing
medicine's memory.
Every clinic is racing to adopt ambient scribes for the hours they give back, and they do give hours back. But look closely at the safety model underneath, and the whole thing rests on four words: the doctor will catch it. No one is checking, at scale, whether that is actually happening.
The question was never whether AI would document medicine. It already does. The real question is who will be protected when the rules arrive, and who will be exposed.
This is not a distant worry. A provincial Auditor General tested approved AI scribes this year. Most got at least one medication wrong. Nearly half recorded detail that was never discussed. Those were the tools that passed review. The reckoning is not a maybe. It is a matter of timing, and of which side of it you are standing on when it comes.