IPCRIME Independent Research | Editorial review: 2026-11-21. Mandiant documented a cluster tracked as UNC5537 using previously stolen Snowflake customer credentials to export data and pressure affected organizations.
Timeline and source record
Mandiant described investigating an initial customer instance in April 2024 and working with Snowflake to notify other customers by late May. It published a detailed report June 10. The findings focus on compromised customer accounts rather than exploitation of Snowflake corporate systems.
What the evidence actually establishes
The research tied observed intrusions to credentials stolen through infostealer activity, often without multifactor authentication or network allow lists. Mandiant explicitly stated it had not found evidence that an intrusion into Snowflake’s enterprise environment caused the customer incidents. Customer data was exfiltrated in investigated cases, but the total victim count depends on the underlying investigation.
Why it matters for affected people
Organizations running cloud data warehouses should review long-lived credentials, contractor devices, application sessions, unusual export volumes and recovery controls. The fact that a stolen credential appears in an infostealer collection does not automatically establish a successful connection at every organization.
What remains to verify
Attribution of all related online sale claims to the same human operator remains an analytical assessment. Exact records and affected persons differ among individual organizations; avoid adding separate company estimates into one unsupported total.
What readers and defenders should take away
- Distinguish a claimed number of records from a confirmed count of unique people or victims.
- Verify any urgent notice through an established company or provider contact; leaked details can make impersonation convincing.
- Do not download or redistribute alleged stolen datasets in an attempt to check the claim.
Method: Research-grounded summary of the source documents linked below; not an independent compromise investigation, live threat-feed observation or association with enforcement agencies. Incident claims and public allegations retain their source attribution.