Purportedly AI-Assisted Hospital-Bed Prioritization Allegedly Delayed Intensive Care Before a Patient's Death in Minas Gerais, Brazil

The family of a 32-year-old patient in Minas Gerais, Brazil, alleged that Core-MG, an AI-assisted hospital-bed allocation system, understated her condition and delayed intensive care. Her sister said the severity score impeded reprioritization. The patient died hours after transfer. The state health department denied a system failure in her case and said physicians assessed patients according to clinical need and bed availability.

Risk classification

• Primary risk domain: 7 AI system safety, failures, & limitations
• Primary risk subdomain: 7.3 Lack of capability or robustness

The Core-MG system failed to perform reliably by misclassifying a critically ill patient's severity, leading to a delayed transfer.

Additional risk subdomains

• 5.2 Loss of human agency and autonomy: The system reportedly stripped doctors of their autonomy to override the AI-generated severity score, preventing them from prioritizing a critical patient.

Causal factors

• Entity: AI
• Intent: Unintentional
• Timing: Post-deployment

The incident was caused by the Core-MG AI system's scoring algorithm failing to accurately reflect the patient's clinical severity, which was an unintended failure of the deployed system.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The system is an 'AI used in healthcare' and critical public infrastructure, specifically for regulating hospital beds and prioritizing emergency patient transfers.

AI system and alleged parties

• AI system: Core-MG
• AI purpose: Resource Allocation; Medical Diagnosis Support
• Behaviour type: Assistant
• Alleged developer: Laboratório de Inovação Tecnológica em Saúde (LAIS/UFRN), AI-enabled healthcare system developers
• Alleged deployer: Minas Gerais State Health Department (SES-MG), Hospitals, Governments, Government of Brazil, Government agencies
• Alleged harmed parties: Women and girls, Women, Rebeca Cardoso Tenente Molina, Patients, Emergency medical patients

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Negligible. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Substantial
• Infrastructure: direct Negligible, indirect Substantial
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Minor
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Minor
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Minor
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Physical

Reported: The report explicitly describes one fatality caused indirectly by the delayed transfer.

Indirectly caused: Rebeca, a 32-year-old patient, died of septic shock after a five-day delay in transferring her to an ICU bed, which her family attributes to the Core-MG system's low severity score.

Inferred additional harm: The report mentions 'other deaths that have been happening since the system was implanted', suggesting there may be additional unquantified fatalities due to system delays.

Infrastructure

Reported: The report describes operational instabilities and overload in the regional healthcare infrastructure.

Indirectly caused: The transition to Core-MG caused operational instabilities, compromised municipal financial planning, and forced Juiz de Fora to receive over 700 patients through 'parallel regulation', overloading the local health network.

Financial

Reported: The report mentions municipal financial planning impacts and the family hiring a private air ambulance.

Indirectly caused: The victim's family had to organize a private air ambulance to transport her, and the municipality of Juiz de Fora reported negative financial impacts on its health system.

Inferred additional harm: The systemic delays likely caused unquantified financial losses for other families forced to seek private care or transport.

Civil rights

Reported: The report describes legal actions taken to protect the population's right to health.

Indirectly caused: The municipality and the Public Prosecutor's Office filed public civil actions to guarantee the safety and continuity of healthcare flows, which were compromised by the system.

Inferred additional harm: The right to life and health of numerous other patients in the region was likely compromised by the system's operational failures.

Psychological

Reported: The report describes severe emotional distress and grief experienced by the victim's family.

Indirectly caused: Rebeca's twin sister and other family members expressed deep grief and 'revolt' over watching her condition worsen while the system failed to recognize the urgency.

Inferred additional harm: Widespread anxiety and distress among local healthcare workers and families of other delayed patients are highly likely given the systemic operational failures.

People affected

• Occurrences reported: 1
• People reportedly harmed: 1
• People reportedly exposed: 701

Potential causes

Management

• Rushed System Deployment: Management deployed Core-MG without validating regional operational readiness.
• Poor Transition Planning: Inefficient management of data migration and risk mitigation during rollout.

Technology

• Rigid Priority Algorithm: Algorithm assigned low score and restricted manual severity overrides.
• Centralized Bed Allocation: Centralized system failed to account for regional distances and transport.
• Operational Instabilities: Platform transition caused system errors and data migration failures.

Data Inputs

• Flawed Patient Scoring Data: Patient severity score of 6.8 did not reflect critical clinical state.
• Incomplete Regional Context: System inputs lacked geographic, logistical, and transport capability data.
• Inadequate Data Migration: Transition occurred without automated or proper data migration from SUSFacil.

Human Factors

• Loss of Clinical Autonomy: Doctors could not override system scores to prioritize urgent transfer.
• Lack of Professional Training: System was implemented without sufficient training for healthcare staff.

Process and Methods

• Sudden System Transition: Abrupt switch to Core/MG without a gradual, safe transition phase.
• Inflexible Prioritization Process: Rigid queue rules prioritized patients with minor score differences over her.
• Centralized Coordination Model: Shifted coordination from macro-regions to centralized Belo Horizonte.

Regulatory Environment

• Overruled Judicial Suspensions: Government overrode judicial decisions to suspend the unstable Core system.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide clear, detailed accounts of the patient's death, the family's specific complaints about the AI scoring system, and the broader systemic issues verified by municipal lawsuits. While the technical details of the algorithm are not fully explained, the operational impact and the role of the AI are clearly documented.
• Ambiguities identified: The exact algorithmic criteria Core-MG uses to calculate the severity score are not detailed, and there is a conflict between the state's claim that doctors retain full control and the family's claim that doctors lost autonomy.
• Alternative interpretations: The incident could be interpreted as a general administrative/IT transition failure rather than an AI-specific failure, but the explicit use of AI for patient prioritization makes the AI safety classification highly appropriate.

The failure of the Core-MG AI bed regulation system in Brazil led to a patient's death and disrupted regional healthcare infrastructure. While highlighting significant risks in deploying AI for critical public services and the danger of reducing human clinical autonomy, the incident remains a localized administrative and public health crisis with minor national security implications.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: Brazil
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. Represents an ongoing strategic concern regarding the integration of AI in public service delivery and critical infrastructure rather than an active crisis.
• Autonomy: Human-supervised. The system was designed to assist human regulators, though operational rigidity reportedly restricted doctors' ability to manually override AI-generated scores.
• Novelty: Evolved capability. Represents an evolution of automated triage errors, transitioning from simple database/IT glitches to complex AI-driven priority scoring systems.

Impact by dimension

• Physical security: Minor. Operational instability in public healthcare infrastructure (bed regulation) led to a patient death and overloaded local networks, but remains localized and manageable through municipal legal actions.
• Information security: Negligible. No information warfare, intelligence compromise, or systematic information manipulation occurred in this incident.
• Sovereignty: Minor. Disrupted local government health administration and prompted lawsuits between municipal and state authorities, but did not threaten national sovereignty or core constitutional processes.
• Economic security: Negligible. Negligible impact on national economic or technological security, restricted to localized municipal health budget strains and private transport costs.
• Societal stability: Minor. Compromised individual right to health and caused public distress, leading to intervention by the Public Prosecutor's Office, but did not trigger widespread civil unrest.