1 CRIT · 8 HIGH · 2 MED · 1 INFO · THREAT RED · 12 items · Generated in 244s
A patch bypass enables unauthenticated remote code execution on default Flowise deployments via npm environment variables.
Tenant-scoped reads can leak another user's data when namespace prefixes overlap in affected LangGraph stores.
Any authenticated user can use server-side fetches to access internal services or cloud metadata and read the responses.
Authenticated users can abuse MLflow AI Gateway for SSRF to reach internal services and possibly steal cloud metadata credentials.
Revoked users can still trigger server-side image generation, bypassing intended access controls and policy enforcement.
Read-only users can obtain server-side tool source and exposed secrets from shared tools.
Authenticated users can make the browser reach internal services and leak returned data into search or RAG output.
Local command injection in an Ollama MCP package can let a user run unintended commands on the host.
Email content can escape UI boundaries and steal tokens or passwords across major webmail providers.
General cybercrime trend headline with no specific AI exploit, vulnerability, or defensive action to map.
AI-made fixes often fail or add flaws, so teams need review and testing before deployment.
Shows AI agents can cause real-world impact when test controls are weak or misconfigured.