Wiz says AI agents ran multi-cloud intrusions end to end
Original title: Loud, Fast, and Everywhere: Anatomy of a Real AI-Driven Cloud Attack
One confirmed entry was Java remote code execution in an exposed Kafka UI, followed by malware and command-and-control.
Wiz's Customer Incident Response Team says recent intrusions were run by AI agents that found entry points, harvested credentials, pivoted across platforms, and rewrote failing scripts. In one multi-cloud case, the agent probed internet-facing services, used a Java remote code execution flaw in an exposed Kafka UI, deployed malware, and opened a command-and-control channel. Activity spanned cloud providers, Kubernetes, CI/CD, identity, secrets, and data stores, and sensitive data was accessed and moved out.
Unconfirmed: The excerpt names no CVE, victim, or threat group, and it cuts off before response guidance.
- Wiz says activity can reach hundreds or thousands of actions per hour.
- The agent worked several platforms at once from the same source.
- Artifacts used labels such as pentest, probe, recon, redteam, and exfil.
- Related resources often shared a numeric task identifier, with letter suffixes on retries.
- Recovered code had descriptive headers and numbered step comments.
Impact: Remote code execution, stolen credentials, and sensitive data access
| Product | Affected |
|---|---|
| Kafka UI | — |
Indicators of compromise
auto-extracted · verify before use · export allAll| Type | Indicator | Context |
|---|---|---|
| domain | webhook.site | vironment. It posted results to the request-catcher service webhook[.]site , committed dumped data to a repository in the organizati |
From the original article
12 min read at wiz.io
Wiz CIRT responded to AI-driven attacks spanning cloud, Kubernetes, CI/CD, and identity. Learn how these attacks behave and how to detect them.
By Eden Abergil · wiz.io
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.