In the wild
Realized harm
Misconfigured Elasticsearch in Serviceaide's agentic AI IT-management platform exposed 483,000 Catholic Health patients' records
AIR-2025-0033 AIR record · 2025-05-serviceaide-catholic-health-elasticsearch-exposure
A real party was affected. This incident caused actual consequences, not a demonstrated capability.
From the supporting source
Serviceaide, a provider of agentic artificial intelligence-based IT management and workflow software, reported to regulators that an inadvertent exposure of data on the web has affected more than 483,000 patients of client Catholic Health, a network of six hospitals and dozens of other facilities in western New York.cybernews.com
The record
Disclosed
2025-05-09
Occurred
2024-11-15
Vendor
Serviceaide / Catholic Health
Product
Serviceaide (agentic AI IT service management)
Risk domain
Privacy › Sensitive data exposure
Also touches
Credential theft
System type
Generative model
Who acted
AI was the target — a misconfigured Elasticsearch index exposed patient records; no model attacked
Attack surface
Agent framework
Autonomy level
Semi-autonomous
Initial vector
Misconfigured datastore
Impact type
Privacy violation, Information disclosure, Credential exposure
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI06 – Memory & Context Poisoning
- MITRE ATLAS
- AML.T0057 – LLM Data Leakage
- NIST AI RMF
- MEASURE 2.10
- EU AI Act
- Article 10
- ISO/IEC
- TR 24028; 23894
- AIUC-1
- A. Data & Privacy
- Scope
- tool misuse
- Mitigation
- Use data anonymization; Implement log sanitization to scrub PII.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
483,000 patients of a six-hospital western New York network; medical records, insurance details and login credentials
Capability abused
Defense & remediation
Recorded defense
access controls on the datastore
Guardrail kinds
Unknown
Guardrail outcome
Silent
Fix shipped
exposure closed; reported to HHS as unauthorised access/disclosure; individuals notified
Taxonomy & references
Curator notes
No model was attacked. The agent platform needed a searchable index over patient records to function, and the index was reachable from the internet. Retrieval infrastructure is the soft underbelly of agentic deployments: it concentrates exactly the data the agent needs and inherits none of the model's guardrails. Discovered 2024-11-15, reported to HHS 2025-05-09. AI Incident Database incident 1070.
Supporting sources
Cite as
Enkrypt AI. "Misconfigured Elasticsearch in Serviceaide's agentic AI IT-management platform exposed 483,000 Catholic Health patients' records". AIR, AIR-2025-0033. https://agentincident.vercel.app/i/AIR-2025-0033/ (accessed 2026-10-19).