Return to Threats

AI Agent Security Incident Log: Breaches, CVEs and Supply Chain

Ajay K 2026-11-12 AI supply chain Critical

Executive Summary

Reported incidents include AI-agent intrusions, attacks involving agent frameworks and coding agents, and a malicious LiteLLM release that distributed credential-stealing malware. Teams operating AI agents, coding agents, or related infrastructure should prioritize dependency and supply-chain controls.

What Happened

The incident log documents reported AI-agent intrusions, evaluation runs reaching real systems, and attacks involving agent frameworks and coding agents. It also records a malicious LiteLLM release that distributed credential-stealing malware through package and container repositories.

Affected Products

  • LiteLLM
  • AI agent frameworks
  • coding agents

Vulnerability / Attack Technique

Malicious package and container release

Technical Mechanism

The summary reports that a malicious LiteLLM release distributed credential-stealing malware through package and container repositories. It also reports intrusions and evaluation runs reaching real systems, but does not provide further technical details on those events.

AI Security Category

LLM03:2025 Supply Chain

Part of the AI Supply Chain and AI SBOM topic.

Relevant Frameworks

LLM03:2025 Supply ChainLLM06:2025 Excessive AgencyMITRE ATLAS: Credential Access

RealGround Analysis

The combination of agent intrusions and a malicious dependency release creates risk across both agent behavior and the software supply chain. Teams should maintain an inventory of agent components, libraries, and containers, while limiting agent access to real systems and protecting credentials from compromised packages.

Recommended Controls

  • Maintain an inventory and SBOM covering AI agents, agent frameworks, libraries, package dependencies, and container images.
  • Pin and verify dependency versions, use trusted registries, and review package and container provenance before deployment.
  • Scan packages and container images for malware and credential-stealing behavior before they enter production.
  • Use least-privilege credentials and isolated execution environments for AI agents, coding agents, and evaluation runs.
  • Monitor agent activity and dependency changes, with alerts for unexpected access to real systems or credential stores.

Related RealGround Service

AI Agent SBOM and LLM Mapping · or start with the free AI Agent Security Scanner.

Sources

Ajay K: https://www.ajayk.sh/blog/ai-agent-security-incident-log/

Summary and analysis written by RealGround from the source above; last updated 2026-11-13.

Talk to AI CISO