HIGH
GHSASUPPLY-CHAIN
MCP-for-Stata: Stata Command Injection via Unsanitized `package` in `ado_package_install`
Unsanitized input in a default MCP tool can trigger OS-level command execution as the Stata-MCP server user.
2 CRIT · 4 HIGH · 5 MED · 1 INFO · THREAT RED · 12 items · Generated in 239s
Unsanitized input in a default MCP tool can trigger OS-level command execution as the Stata-MCP server user.
Tenant-scoped reads can leak another tenant's stored data when namespace prefixes overlap.
Path traversal in Triton can lead to code execution on inference hosts, risking model service compromise.
Absolute path traversal in Triton may enable code execution and disclose sensitive data on affected Linux servers.
Unauthenticated path traversal can expose session log files and leak sensitive data from the server filesystem.
A path traversal bug in Triton Inference Server could let an attacker crash the service and disrupt model availability.
Improper input validation in Triton Inference Server could let an attacker trigger denial of service.
A flaw in Triton can exhaust resources and disrupt inference service availability.
An actively exploited Ray RCE could let attackers compromise AI infrastructure and reach private corporate networks.
Actively exploited Ray flaw can lead to browser-based remote code execution on AI/ML infrastructure.
Funding news on AI governance tooling, not a disclosed vulnerability or active threat.
Shows an AI assistant can be manipulated to disclose internal architecture and security details attackers can reuse.