The vulnerability arises from an oversight in the unparsed opaque token validation path. The validateOpaqueToken function interprets the active field from the OAuth 2.0 introspection response as a pointer to a boolean. The code only rejects a token when the response explicitly contains an active field set to false; if the active key is omitted, the pointer remains nil and the check short‑circuits. Consequently, the toolbox accepts tokens that lack the required active claim, enabling an attacker to use any such token to gain access to protected tools and data.

Affected systems are implementations of Google MCP Toolbox for Databases (googleapis/mcp-toolbox). The issue applies to any deployment that uses the validateOpaqueToken path to authenticate users through an OAuth 2.0 introspection endpoint as defined in RFC 7662. The specific version ranges affected are not detailed, so all current releases prior to the fix should be considered vulnerable unless explicitly updated.

The CVSS base score of 9.3 indicates critical severity, with the potential for remote attackers to compromise sensitive data. The EPSS score is unavailable but the vulnerability has not been listed in the CISA KEV catalog. The attack requires access to an introspection endpoint that may return a response missing the active field, which could be manipulated by a malicious or compromised OAuth provider. The flaw is exploitable over the network and does not necessitate privileged local access, implying a high likelihood of exploitation in environments where such introspection endpoints are reachable from the internet or compromised.