In many cases, patients first notice something is off because the record reads differently than they expected—generated summaries, automated measurements, “decision support” language, or references to systems used in imaging, planning, or workflow documentation.
That can matter legally when:
- the AI output influenced decisions without appropriate clinical verification,
- documentation was incomplete, inconsistent, or overly dependent on automated text,
- clinicians failed to recognize a warning sign or abnormal result,
- or the workflow didn’t meet the safety expectations of a reasonable provider.
Important: AI doesn’t automatically mean negligence. But it can create new ways errors happen—and it can also create new evidence you may not want to lose.


