TYPO3 AI Integration EUVD-2026-104863

Exploitation requires three conditions all confirmed by the advisory: (1) the TYPO3 AI integration extension must be installed on TYPO3 10 or 11, (2) an OpenAI API key must be configured - the advisory explicitly states the issue 'only applies if an API key has been configured,' and (3) the attacker must already hold valid authenticated backend credentials (PR:L), even at low/editor privilege. The key is embedded in every backend page by design so the browser-side editor plugin can call OpenAI directly, so no special page or feature toggle beyond key configuration is needed. Limiting factors: no exposure on default installs without a configured key, no exposure to unauthenticated or frontend-only users, and impact is confined to disclosure of the API key (no integrity or availability effect).

This is a moderate, credential-disclosure risk rather than a critical remote-code-execution threat. The vendor-supplied CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N/VC:H) scores 7.1 and correctly reflects that exploitation requires an already-authenticated backend user (PR:L) but no user interaction and low complexity; the only confidentiality impact is disclosure of the configured OpenAI API key (VC:H, VI:N, VA:N). No EPSS score, CISA KEV listing, or public POC was provided in the intelligence, so there is no public exploit identified at time of analysis and no evidence of active exploitation. The practical risk is insider/abuse-of-access: any backend editor can harvest the key and run up charges on the operator's OpenAI account, or pivot the key for use elsewhere. The CVSS signals are internally consistent (network-reachable backend, authenticated, confidentiality-only), so there is no AV-versus-description conflict to flag. Net priority is medium - contingent entirely on the extension being installed AND an API key being configured; absent that configuration there is no exposure.

A low-privileged but authenticated TYPO3 backend editor opens any backend page and views the page source or network traffic, where the OpenAI API key is embedded so the editor plugin can call OpenAI from the browser. The editor copies the key and uses it against OpenAI directly, billing usage to the site operator. No public exploit code is required - the key is readable with ordinary browser developer tools.