Starting with release 2.30.0, the AI Bridge proxy endpoints authenticate requests through a server‑side authorization check that validates key format, expiration, and whether the key has been deleted or belongs to a system user, but it does not examine the suspension status of the account. Consequently, a suspended user who still possesses an unexpired token can continue to invoke the AI Bridge LLM proxy endpoints until that key is revoked or expires. The flaw does not allow arbitrary code execution or direct data exfiltration beyond the normal use of the service; it merely permits continued use of the AI Bridge by a user that the organization intends to block.

The issue affects the coder:coder product. Versions 2.30.0 through 2.32.6, 2.33.7, and 2.34.1 are impacted. Fixed releases are 2.32.7, 2.33.8, and 2.34.2. Users running any of the affected releases should upgrade to a patched version to eliminate the flaw.

The CVSS score of 5.4 signals moderate severity, and the EPSS score of less than 1% indicates a very low exploitation probability at present. The vulnerability is not listed in CISA’s KEV catalog, suggesting no widespread exploitation has been documented. Exploitation requires possession of a suspended user’s unexpired API key; an attacker who learns or intercepts such a key can continue to call the AI Bridge endpoints until the key is rotated or deleted. The flaw is a broken access control scenario (CWE‑863) and is mitigated by enforcing suspension checks during authentication.