Vulnerabilities Threaten to Break Chainlit AI Framework

Vulnerabilities Threaten to Break Chainlit AI Framework reports two high‑severity flaws in the Chainlit open‑source framework for creating conversational chatbots, with the team at Zafran Security noting the risks extend to cloud environments. The first flaw, CVE‑2025‑22218 (CVSS 7.7), stems from Chainlit’s ability to attach files to messages; via a specific API endpoint an attacker can update these elements to point to arbitrary server files, which may then be downloaded from the attacker’s session.

The second vulnerability, a server‑side request forgery (SSRF) tracked as CVE‑2026‑21219 (CVSS 7.0), can trigger when a URL is supplied in a custom element and SQLAlchemy is used for data persistence, allowing the server to fetch restricted resources. Researchers warned that exploiting one flaw to learn internal secrets and then using the other to reach sensitive systems could enable cloud account takeovers, data theft, or even source‑code theft, particularly if Chainlit runs on AWS EC2 with IMDSv1 enabled.

According to Dark Reading, Chainlit released version 2.9.4 last month to fix the vulnerabilities, which were described as affecting millions of users. 20 January 2026.