The vulnerability is a classic cross‑site scripting flaw that allows an attacker to inject javascript: links into the chat content. Because Deep Chat’s RemarkableConfig.createNew disables link validation, these links can be embedded in responses, user messages, or chat history. When a user clicks such a link in the embedding web page, the malicious script runs with the page’s privileges, potentially exfiltrating data, hijacking the session, or re‑rendering deceptive content.

The flaw affects the Deep Chat application, version 2.5.1, distributed by OvidijusParsiunas. The affected software is listed under the CPE cpe:2.3:a:thinkinai:deepchat:*:*:*:*:*:*:*:*. The issue is specific to this version; later releases may have included a fix.

The CVSS score of 2.1 classifies the issue as low severity. No EPSS score is available and the vulnerability is not listed in the CISA KEV catalog. Attackers could exploit this by crafting malicious Markdown links either in AI-generated responses, user‑submitted content, or pre‑loaded chat history; the need to click the link means the attack requires user interaction. Because the script runs in the context of the host page, the impact can range from data leakage to full session compromise, but actual exploitation likelihood remains low in the absence of a public exploit.