Vulnerabilities Threaten to Break Chainlit AI Framework

Researchers at Zafran disclosed two high-severity flaws in the Chainlit conversational AI framework: one lets an attacker update message "custom elements" to read arbitrary server files, and the other is a server-side request forgery (SSRF) that can fetch internal URLs (including AWS IMDSv1), enabling credential theft, cloud takeovers, and data/source-code exfiltration; Chainlit released version 2.9.4 to patch the issues.