INCIDENT
MCP - AI security threats have expanded beyond just the
AI security threats have expanded beyond just the model, including vulnerabilities in agent configurations, MCP servers, and other supporting infrastructure that standard tools may not be able to handle.
Updated: 3/13/2026
high Severity
Status: active
Description
AI security shifted beyond the model itself. Your Cursor rules file might hide instructions. Your MCP server could be poisoned. Agent configs leak credentials. The threat model expanded past what standard tools can handle.
Impact
AI security threats have expanded beyond just the model, including vulnerabilities in agent configurations, MCP servers, and other supporting infrastructure that standard tools may not be able to handle.
Attack Vectors
- security
- agent configuration