Google Application Integration CVE-2026-81375

Requires an authenticated principal with access to the EmailTask component of Google Cloud Application Integration (PR:L) able to define/edit an integration task and supply a crafted attachment file path; no user interaction is needed. Exploitation is limited to versions prior to 2026-06-30 - because this is a Google-managed SaaS service, the 2026-06-30 server-side patch is applied universally and no customer-side vulnerable configuration persists, so no self-hosted or unpatched instances remain exploitable.

The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N/VC:H/SC:H, base 8.3) describes a network-reachable, low-complexity confused-deputy flaw requiring only low-privilege authentication and no user interaction, with high confidentiality impact on both the vulnerable and a subsequent system (SC:H) - consistent with the service's identity boundary being abused to read Google-internal files. However, several risk-reducing signals dominate: there is no EPSS score, no CISA KEV listing, and no public POC in the provided data, so this is not confirmed exploited and no exploit code is known. Most importantly, Google states the flaw was patched server-side on 2026-06-30 with no customer action required - because Application Integration is a fully-managed GCP SaaS service, the fix is universally deployed and no vulnerable instances remain in customer control. Net: a genuinely serious confidentiality/confused-deputy issue in design terms, but effective real-world risk today is low-to-negligible because it is fully remediated at the provider and was never shown to be exploited. No conflicting signals in the data, though the description does not enumerate exactly which 'Google-internal files' were reachable, so the concrete blast radius is vendor-asserted rather than independently verifiable.

An authenticated Google Cloud customer with access to Application Integration configures an EmailTask and supplies a crafted attachment file path that points at Google-internal resources rather than the customer's own data. The service, acting as a confused deputy with its own elevated privileges, reads and attaches those internal files, exfiltrating them to the attacker. No public exploit code is identified at time of analysis, and the flaw is patched server-side as of 2026-06-30.