Blab Tech
AI Drug-Monitoring System Missed Fentanyl Theft at Tennessee Hospital, Prompting Calls for Human Oversight
When a nurse’s clandestine fentanyl theft slipped through the cracks of an AI monitoring system, the incident rattled hospitals across the country. In 2025, a nurse anesthetist named John Stevenson quietly siphoned leftover fentanyl from Erlanger Hospital’s anesthesia supply room in Chattanooga, Tennessee, for nearly four months. The theft was finally uncovered when anesthesia staff observed Stevenson’s erratic behavior during a surgery shift and reported the suspicious activity to hospital security.
Erlanger’s medication‑tracking software, Sentri7, is installed in hundreds of U.S. hospitals and is marketed as a tool that can spot missing drugs faster than human staff. According to the hospital’s internal investigation, Sentri7 failed to flag Stevenson’s repeated diversion of fentanyl—a potent opioid used for surgical and post‑operative pain management. The compliance officer reviewed the system’s logs and found no alerts related to the theft.
The Tennessee Department of Health and Health Secretary Robert F. Kennedy Jr. were approached by local media to determine whether state regulators will mandate human oversight of AI‑based drug‑diversion monitoring. The Department has no current policy requiring such oversight. However, doctors at Johns Hopkins Hospital have noted that in practice, human reviewers typically supervise similar technology, even though that arrangement is not legally required.
Fentanyl’s extreme potency—50 to 100 times stronger than morphine—makes it a frequent target for diversion. The Drug Enforcement Administration requires hospitals to report lost or stolen drugs confidentially, and the failure of Sentri7 to detect Stevenson’s theft raises questions about the reliability of AI systems in high‑stakes environments.
Sentri7’s developers claim the platform uses machine‑learning algorithms to analyze medication inventory data in real time, generating alerts when usage patterns deviate from expectations. In Stevenson’s case, the software did not trigger a warning, suggesting a gap in its detection logic or in the data fed to it.
The incident is not isolated. A 2026 report from the Center for Health AI indicates that several hospitals have experienced similar lapses, prompting a broader debate about the role of AI in patient safety. Some clinicians argue that AI should serve as an assistant, not a replacement for human judgment, especially when dealing with drug diversion.
Erlanger fired Stevenson and reported the theft to the Tennessee Department of Health. The state records confirm that the hospital’s compliance officer reviewed Sentri7’s logs and found no alerts related to the theft.
Calls for clearer regulatory guidance are mounting. Advocates for patient safety contend that hospitals should be required to conduct regular audits of AI monitoring systems and maintain human oversight. Meanwhile, developers of Sentri7 and similar platforms emphasize that AI can process vast amounts of data more quickly than humans, but they also acknowledge that human review remains essential.
As hospitals continue to adopt AI tools for drug‑diversion detection, this incident underscores the need for robust oversight mechanisms. The Tennessee Department of Health has not yet announced any policy changes, but the state’s response will likely influence national standards for AI use in healthcare.
The case remains a cautionary example of how reliance on automated systems can create blind spots in critical safety processes. It also highlights the ongoing tension between technological efficiency and the need for human judgment in safeguarding patient care.
Erlanger’s medication‑tracking software, Sentri7, is installed in hundreds of U.S. hospitals and is marketed as a tool that can spot missing drugs faster than human staff. According to the hospital’s internal investigation, Sentri7 failed to flag Stevenson’s repeated diversion of fentanyl—a potent opioid used for surgical and post‑operative pain management. The compliance officer reviewed the system’s logs and found no alerts related to the theft.
The Tennessee Department of Health and Health Secretary Robert F. Kennedy Jr. were approached by local media to determine whether state regulators will mandate human oversight of AI‑based drug‑diversion monitoring. The Department has no current policy requiring such oversight. However, doctors at Johns Hopkins Hospital have noted that in practice, human reviewers typically supervise similar technology, even though that arrangement is not legally required.
Fentanyl’s extreme potency—50 to 100 times stronger than morphine—makes it a frequent target for diversion. The Drug Enforcement Administration requires hospitals to report lost or stolen drugs confidentially, and the failure of Sentri7 to detect Stevenson’s theft raises questions about the reliability of AI systems in high‑stakes environments.
Sentri7’s developers claim the platform uses machine‑learning algorithms to analyze medication inventory data in real time, generating alerts when usage patterns deviate from expectations. In Stevenson’s case, the software did not trigger a warning, suggesting a gap in its detection logic or in the data fed to it.
The incident is not isolated. A 2026 report from the Center for Health AI indicates that several hospitals have experienced similar lapses, prompting a broader debate about the role of AI in patient safety. Some clinicians argue that AI should serve as an assistant, not a replacement for human judgment, especially when dealing with drug diversion.
Erlanger fired Stevenson and reported the theft to the Tennessee Department of Health. The state records confirm that the hospital’s compliance officer reviewed Sentri7’s logs and found no alerts related to the theft.
Calls for clearer regulatory guidance are mounting. Advocates for patient safety contend that hospitals should be required to conduct regular audits of AI monitoring systems and maintain human oversight. Meanwhile, developers of Sentri7 and similar platforms emphasize that AI can process vast amounts of data more quickly than humans, but they also acknowledge that human review remains essential.
As hospitals continue to adopt AI tools for drug‑diversion detection, this incident underscores the need for robust oversight mechanisms. The Tennessee Department of Health has not yet announced any policy changes, but the state’s response will likely influence national standards for AI use in healthcare.
The case remains a cautionary example of how reliance on automated systems can create blind spots in critical safety processes. It also highlights the ongoing tension between technological efficiency and the need for human judgment in safeguarding patient care.
