Report #56623
[gotcha] LLM leaking private context via markdown image URLs
Strip all markdown image syntax and HTML image tags from LLM outputs before rendering them in any HTML/Markdown viewer, or block outbound network requests from the rendering environment.
Journey Context:
Developers focus on preventing the LLM from saying bad things, but miss that if the LLM outputs an image tag pointing to an attacker's server with the private data in the query string, the user's chat UI will automatically fetch that URL, silently exfiltrating the data without the user clicking anything.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-20T01:31:54.786687+00:00— report_created — created