The global patient safety nonprofit ECRI has expanded its Problem Reporting Network to capture and investigate errors, malfunctions, and near misses involving artificial intelligence tools used in patient care. The organization is asking healthcare providers, health systems, and clinicians nationwide to report any incident where an AI-enabled tool may have contributed to an error or introduced risk into care delivery.
ECRI triages and investigates each submission, then shares findings directly with the person who filed the report. The expansion complements the ECRI and ISMP Patient Safety Organization, which has collected and analyzed more than 8 million safety event reports from healthcare providers across the country.
AI adoption is outpacing oversight
Clinical AI tools and AI-enabled medical devices are being deployed faster than the safeguards designed to prevent, identify, and respond to failures. AI now assists with diagnostic imaging, clinical decision support, and patient-facing chatbots, but no centralized healthcare-specific mechanism tracks how often these tools produce incorrect outputs - or whether those outputs reach a patient.
"ECRI has persistently emphasized the risk of adopting AI with insufficient scrutiny," said Scott Lucas, PhD, ECRI's Vice President of Devices, Therapeutics, and Technology. "Although we appreciate AI's tremendous potential, we don't yet have a clear picture of its downstream impact in healthcare. Without a robust reporting dataset and analysis, the industry cannot sufficiently improve the design and integration of AI tools and devices. We must look to evidence and data to understand the evolving risks and associated system factors, to enable the use of the safest, most effective technologies."
Survey data shows AI errors are happening - and often going unnoticed
In a recent ECRI survey of 124 respondents - mostly quality, safety, risk, and compliance leaders - nearly one-third said they encountered an AI output they believed was incorrect or misleading over the past year. Another 35% were unsure whether they had encountered an error, and 9% reported that an AI error reached a patient or affected a care decision.
Ambient scribes were the most commonly encountered AI in the survey interactions, cited by 37% of respondents. EHR-embedded clinical decision support followed at 31%, tied with clinical LLM assistants and common systems.
ECRI has evaluated AI-enabled devices and tools across imaging, fall prevention, anesthesia, wearable technology for chronic disease management, cardiovascular diagnoses, colonoscopy systems, and behavioral therapy.
How the Problem Reporting Network works
ECRI's Problem Reporting Network has operated since 1972 as a free, confidential channel for reporting medical device and technology problems. Every submission is triaged and investigated by ECRI's clinical and engineering experts, who follow up directly with the submitter to share what they found.
When a safety risk is identified, ECRI issues hazard reports to manufacturers, providers, and regulatory agencies, describing the problem and offering actionable recommendations. The network now explicitly includes a pathway for reporting suspected AI-related errors, incorrect outputs, or unsafe AI behavior encountered in clinical or operational use.
ECRI also encourages healthcare organizations to submit safety reports through all applicable required channels, including state and federal reporting and Patient Safety Organizations.
Why this matters for healthcare professionals
Clinicians and safety leaders are currently making care decisions with AI tools whose failure rates are largely unknown. The 35% of survey respondents who were unsure whether they had encountered an AI error is itself a finding: most healthcare workers have no reliable way to recognize or report when an AI output is wrong. Submitting incidents to ECRI's network gives individual clinicians a direct line to experts who can investigate the issue and share what they find - and the resulting hazard reports give the broader field actionable guidance on which tools carry known risks.
Your membership also unlocks:
