What happened
FlowIntel contains multiple security flaws in its note-variable resolution, chatbot API tooling and AI report generation. References are resolved with the viewing user's privileges, the model can invoke unrestricted endpoints, and the Ollama bearer token plus case content are exposed in the system process table. CWE-200/CWE-284/CWE-862, CVSS 4.0 7.1.
Why it matters
The chatbot API tools expose data-modification and restricted endpoints to a prompt-influenced LLM, and the AI report path leaks the model's bearer token and full case content to any local user — combining authorization bypass with credential leakage in an AI investigation tool.
Attack vector
Note-variable references are resolved using the viewing user's privileges rather than the destination case's audience; the chatbot API tool lets the LLM invoke any endpoint (including admin routes and state-changing GETs) without an allow-list or path validation; AI report generation passes the Ollama token and full case content as command-line arguments to a child process.
Affected systems
FlowIntel < 4.1.0 (app/case/CaseCore.py, app/case/case_notes.py, app/chatbot/chatbot_core.py, app/utils/note_variables.py)
Mitigation
Upgrade FlowIntel to 4.1.0 or later.