Blab Tech
AI Missed 5 Fentanyl Diversion Warnings, Tennessee Nurse Fired for Theft
A Tennessee hospital’s artificial‑intelligence (AI) drug‑diversion system failed to flag five red‑flag alerts, allowing a nurse to siphon fentanyl before she was fired for theft. State records released by the Tennessee Nursing Board show that John Stevenson, a certified registered nurse anesthetist at Erlanger Hospital in Chattanooga, was terminated after admitting to taking leftover fentanyl from the hospital’s supply.
Stevenson first attracted suspicion after a failed drug test and reports from coworkers that he appeared impaired. When investigators asked him about the missing medication, he confessed to stealing it. The Nursing Board’s findings confirm that the AI system, which monitors prescription patterns for potential diversion, did not trigger any alerts related to his activity.
Drug diversion—where legally prescribed controlled substances are redirected from intended patients to other parties—remains a persistent problem in health care. Fentanyl, a synthetic opioid up to 50 times stronger than heroin, is a frequent target because of its high potency and value on the illicit market.
Erlanger Hospital, a non‑profit system that serves a 50,000‑square‑mile region of East Tennessee, North Georgia, North Alabama and western North Carolina, relies on AI to help identify potential diversion. According to the Nursing Board report, the system failed to raise any alarms for Stevenson’s behavior, a lapse that prompted the hospital to terminate him.
Hospitals that also use AI for medication safety note that human oversight remains integral. A spokesperson for Johns Hopkins Hospital, which employs a similar system, said, “We have humans reviewing the AI outputs, but that is not mandated by law.” The same officials emphasized that AI should act as an assistant, with clinicians making the final decisions.
The incident drew attention from local media and raised questions about state oversight of AI tools in hospitals. FOX 17 News’ Johnny Maffei contacted the Tennessee Department of Health and Health Secretary Robert F. Kennedy Jr. to ask whether the state would consider mandating human review of AI‑generated diversion alerts. The Secretary’s office has not yet issued a response, and the Department of Health says it is monitoring the situation but has not announced any policy changes.
The case highlights the limits of current AI systems in detecting subtle patterns of drug diversion. While AI can process large volumes of data faster than humans, it can still miss critical signals, especially when those signals are rare or when the system’s training data is incomplete.
Erlanger Hospital has not released a statement beyond the Nursing Board’s findings, and the director of medication safety—who was not named in the report—has not commented.
The Tennessee Nursing Board’s findings are part of a broader national conversation about the role of AI in healthcare safety. Several states are reviewing their regulatory frameworks for AI tools that support clinical decision‑making.
At present, the only confirmed outcome is that John Stevenson was terminated and that the AI system failed to flag five diversion warnings. No further legal action or regulatory changes have been announced.
The incident serves as a cautionary example for hospitals that rely on AI for drug‑diversion detection and underscores the need for clear policies that require human oversight of AI outputs, especially in high‑stakes environments such as medication safety.
The Tennessee Department of Health and the state’s nursing board will likely continue to examine the incident as part of ongoing efforts to strengthen drug‑diversion safeguards. The focus remains on ensuring that AI tools are used as aids rather than replacements for human judgment in critical safety functions.
The situation will be monitored for any future regulatory announcements or policy changes that may arise from this incident.
Stevenson first attracted suspicion after a failed drug test and reports from coworkers that he appeared impaired. When investigators asked him about the missing medication, he confessed to stealing it. The Nursing Board’s findings confirm that the AI system, which monitors prescription patterns for potential diversion, did not trigger any alerts related to his activity.
Drug diversion—where legally prescribed controlled substances are redirected from intended patients to other parties—remains a persistent problem in health care. Fentanyl, a synthetic opioid up to 50 times stronger than heroin, is a frequent target because of its high potency and value on the illicit market.
Erlanger Hospital, a non‑profit system that serves a 50,000‑square‑mile region of East Tennessee, North Georgia, North Alabama and western North Carolina, relies on AI to help identify potential diversion. According to the Nursing Board report, the system failed to raise any alarms for Stevenson’s behavior, a lapse that prompted the hospital to terminate him.
Hospitals that also use AI for medication safety note that human oversight remains integral. A spokesperson for Johns Hopkins Hospital, which employs a similar system, said, “We have humans reviewing the AI outputs, but that is not mandated by law.” The same officials emphasized that AI should act as an assistant, with clinicians making the final decisions.
The incident drew attention from local media and raised questions about state oversight of AI tools in hospitals. FOX 17 News’ Johnny Maffei contacted the Tennessee Department of Health and Health Secretary Robert F. Kennedy Jr. to ask whether the state would consider mandating human review of AI‑generated diversion alerts. The Secretary’s office has not yet issued a response, and the Department of Health says it is monitoring the situation but has not announced any policy changes.
The case highlights the limits of current AI systems in detecting subtle patterns of drug diversion. While AI can process large volumes of data faster than humans, it can still miss critical signals, especially when those signals are rare or when the system’s training data is incomplete.
Erlanger Hospital has not released a statement beyond the Nursing Board’s findings, and the director of medication safety—who was not named in the report—has not commented.
The Tennessee Nursing Board’s findings are part of a broader national conversation about the role of AI in healthcare safety. Several states are reviewing their regulatory frameworks for AI tools that support clinical decision‑making.
At present, the only confirmed outcome is that John Stevenson was terminated and that the AI system failed to flag five diversion warnings. No further legal action or regulatory changes have been announced.
The incident serves as a cautionary example for hospitals that rely on AI for drug‑diversion detection and underscores the need for clear policies that require human oversight of AI outputs, especially in high‑stakes environments such as medication safety.
The Tennessee Department of Health and the state’s nursing board will likely continue to examine the incident as part of ongoing efforts to strengthen drug‑diversion safeguards. The focus remains on ensuring that AI tools are used as aids rather than replacements for human judgment in critical safety functions.
The situation will be monitored for any future regulatory announcements or policy changes that may arise from this incident.
