EXECUTIVE SUMMARY:
CVE-2026-55435 with a CVSS score of 5.4 is an authorization bypass vulnerability affecting Coder v2 versions 2.30.0 through 2.34.1, specifically within the 2.32, 2.33, and 2.34 release lines. This technical flaw arises because the AI Bridge proxy authentication mechanism validates API key format, expiry, and deletion status but fails to verify if the associated account is suspended. An attacker can exploit this vulnerability by utilizing a previously issued, unexpired API key belonging to a suspended user, thereby bypassing intended access controls. This capability allows the attacker to maintain persistent access to AI Bridge LLM proxy endpoints, enabling them to consume paid provider resources billed to the deployment and invoke injected Model Context Protocol tools. The business impact of this exploitation includes direct financial costs from resource consumption and potential security risks involving unauthorized tool execution. Exploitation is strictly conditional upon the attacker possessing a valid API token that was issued prior to the user's suspension, limiting the scope to accounts with existing credentials.[/subscribe_to_unlock_form]
EXECUTIVE SUMMARY:
CVE-2026-55435 with a CVSS score of 5.4 is an authorization bypass vulnerability affecting Coder v2 versions 2.30.0 through 2.34.1, specifically within the 2.32, 2.33, and 2.34 release lines. This technical flaw arises because the AI Bridge proxy authentication mechanism validates API key format, expiry, and deletion status but fails to verify if the associated account is suspended. An attacker can exploit this vulnerability by utilizing a previously issued, unexpired API key belonging to a suspended user, thereby bypassing intended access controls. This capability allows the attacker to maintain persistent access to AI Bridge LLM proxy endpoints, enabling them to consume paid provider resources billed to the deployment and invoke injected Model Context Protocol tools. The business impact of this exploitation includes direct financial costs from resource consumption and potential security risks involving unauthorized tool execution. Exploitation is strictly conditional upon the attacker possessing a valid API token that was issued prior to the user's suspension, limiting the scope to accounts with existing credentials.[emaillocker id="1283"]
RECOMMENDATION:
REFERENCES:
The following reports contain further technical details:
https://github.com/advisories/GHSA-wqxv-w64v-5wh6