Cross-site Scripting via Discourse-ai SharedAiConversation onebox in Discourse

An attacker could exploit this XSS vulnerability by crafting a Discourse AI Bot conversation containing malicious HTML entities. When this conversation is shared and oneboxed into a Discourse post, the injected script will execute in the context of the user viewing the post. This could lead to various malicious actions, including stealing cookies, redirecting users to phishing sites, or defacing the website. The impact is particularly severe because it affects the entire Discourse community viewing the compromised post, potentially impacting a large number of users.

Discourse installations utilizing the Discourse AI plugin, particularly those with public forums or where AI Bot conversations are frequently shared, are at risk. Shared hosting environments running Discourse are also vulnerable, as the plugin's security depends on the host's overall security posture.

• discourse: Check Discourse logs for unusual JavaScript execution or suspicious URL patterns in post content. • generic web: Use curl/wget to inspect the HTML source code of posts that onebox AI Bot conversations for injected scripts.

curl -s 'https://your-discourse-site.com/t/example-post' | grep -i '<script>' 

1. 2025-01-14Disclosure

   disclosure

1. Reserviert2024-11-29
2. Veröffentlicht2025-01-14
3. Geändert2025-01-15
4. EPSS aktualisiert2026-04-02

The primary mitigation for CVE-2024-54142 is to upgrade to Discourse AI version 92f122c or later. This version contains a fix that prevents the leakage of malicious HTML entities. If upgrading is not immediately feasible, a temporary workaround is to remove all groups from the ai bot public sharing allowed groups site setting. This will disable the sharing of AI Bot conversations, effectively preventing the vulnerability from being exploited. After upgrading, confirm the fix by sharing a test conversation with known malicious HTML entities and verifying that the script does not execute.