2 CRIT · 4 HIGH · 5 MED · 1 INFO · THREAT RED · 12 items · Generated in 245s
A patch bypass enables unauthenticated remote code execution on default Flowise deployments via npm_config_yes and npx auto-install.
Scoped reads can leak one tenant's stored data into another tenant's results due to namespace boundary failures.
Any authenticated user may access internal services or cloud metadata through the server, exposing sensitive data and enabling SSRF-style abuse.
Read-only users can view server-side tool code and exposed secrets, enabling further compromise of shared AI tooling.
Revoked users can still trigger server-side image generation, bypassing admin policy and consuming restricted AI resources.
Authenticated local input validation flaw can crash affected vLLM Gaudi components and reduce service availability.
Any logged-in user can change or remove site languages, causing unauthorized configuration changes and possible service disruption.
Low-privilege users can replace the translation API token, disrupting service or redirecting translation traffic.
Unauthenticated RCE on SharePoint can give full server access and admin compromise on exposed enterprise systems.
Broad agent access plus vague tasks can lead to unintended actions and data exposure without clear intent and permission controls.
Attackers can hijack AI agents via manipulated alerts and blocked events, exposing weak identity and control boundaries.
Highlights governance and legal risk from rushed AI adoption, but describes no exploit, breach, or technical weakness.