AI Incidents
Back to incident register
Source reviewedSecurity breachHigh

Google links generative-AI tooling to BREEZE COMET financial intrusions in Brazil

Google Threat Intelligence and Mandiant reported compromises of Brazilian organizations in which BREEZE COMET used large language models to support reconnaissance, credential validation, deployment, victim-specific pivoting and data extraction.

First observed
Sep 1, 2026, 12:00:03 PM
Disclosed
Sep 1, 2026, 12:00:03 PM
Status
Monitoring
Confidence
94%
Organization
BREEZE COMET, Mandiant, Google Threat Intelligence Group
Last reviewed
Sep 4, 2026

Observed

Facts supported by sources

  • Mandiant reported investigating compromises of Brazilian financial, retail and ecommerce organizations attributed to BREEZE COMET.
  • Recovered tooling and scripts indicated use of generative AI and large language models to support reconnaissance, credential validation, mass deployment, victim-specific pivoting and data extraction.
  • The report described hundreds of fraudulent transactions across two waves and at least one theft worth tens of thousands of US dollars. Because individual event dates were not disclosed, this record uses the disclosure date as its timeline anchor.

Assessment

Interpretation, not observation

  • The source supports AI-assisted malware development and operations, not autonomous criminal action.
  • The campaign was described as active and developing at publication time.

Impact

Impact

Financial theft, credential compromise, malware deployment and data extraction affecting organizations in Brazil.

Response

Response

Google and Mandiant published technical indicators, detection guidance and hardening recommendations. Responses by the individual affected organizations were not disclosed.

Methodology: Facts, interpretations, and uncertainty are kept separate. Confidence describes the strength of the evidence, not a probability estimate.