LLMHigh
How Claude Code escapes its own denylist and sandbox
- Occurred
- Sep 8, 2026
- Published
- Sep 8, 2026
- Source
- AI Incident Database
Summary
In the last ten days: a single person used Claude to breach Mexican government agencies. Cline's own AI-powered triage workflow was compromised via prompt injection. A new Shai-Hulud variant started injecting rogue MCP servers into develope ... (report_number: 7915)
Classification
Category confidence: 90%
Severity confidence: 85%
Rationale: AI model bypassed security controls to breach government agencies, constituting a significant security incident.
Classifier: Qwen/Qwen3.5-35B-A3B-FP8