CVE-2026-43900: DeepChat XSS Bypass Threatens AI Implementations

The National Vulnerability Database has issued a critical alert for CVE-2026-43900, affecting DeepChat, an open-source AI agent platform. This Cross-Site Scripting (XSS) vulnerability, rated 9.3 (CRITICAL), stems from a mismatch in validation logic: the backend’s SVG sanitizer (src/main/lib/svgSanitizer.ts) attempts to block javascript: protocols using plain-text regex, but fails to account for HTML entity decoding before Vue’s v-html DOM insertion in the SvgArtifact.vue component.

Attackers can exploit this by injecting SVG artifacts with obfuscated entities (e.g., javascript:alert(1)). This completely bypasses the sanitizer, leading to arbitrary JavaScript execution when a victim interacts with the rendered SVG element. The impact is significant: compromise of user sessions, data exfiltration, or further client-side attacks are all on the table. The vulnerability is fixed in DeepChat v1.0.4-beta.1.

This isn’t just about a single product; it highlights a recurring flaw in security engineering. Relying solely on regex for sanitization, especially without prior HTML entity decoding, is a rookie mistake that continues to plague applications. The attacker’s calculus here is simple: find a parsing discrepancy and weaponize it. Defenders need to assume such bypasses exist and build more robust, multi-layered input validation and sanitization. This vulnerability is a stark reminder that even open-source projects, often assumed to be more secure due to community scrutiny, can harbor critical design flaws.