AI Engine CVE-2026-65545

No authentication is required to deliver the XSS payload, as confirmed by PR:N in the CVSS vector. Exploitation does require victim interaction (UI:R per CVSS): a human user - ideally an authenticated WordPress administrator - must trigger the malicious content, such as clicking a crafted link or visiting a URL with the injected parameter. The specific plugin feature, endpoint, or input field containing the injection point is not disclosed in available intelligence. Exploitation is limited to WordPress installations with the AI Engine plugin installed and activated on versions up to and including 3.6.8; sites not running this plugin are unaffected. The most impactful scenario depends on a privileged (e.g., admin-level) victim interacting with the payload, since the scope change enables actions across the victim's authenticated session.

The CVSS 7.1 score captures real structural risk: network-accessible (AV:N), low complexity (AC:L), no authentication required (PR:N), and scope change (S:C) that enables impact on privileged user sessions. The primary exploitation constraint is mandatory victim interaction (UI:R), meaning the attacker cannot achieve impact passively - a social engineering step is required to deliver the payload to a logged-in user. No EPSS data was available in the provided intelligence, limiting probabilistic exploitation likelihood assessment. The absence from CISA KEV and unconfirmed public exploit code reduce immediate urgency relative to the raw score. However, XSS against WordPress plugins with unauthenticated injection points is a well-understood and frequently exploited attack class; generic exploitation tooling lowers the barrier significantly once a specific site is targeted. For deployments where administrators routinely browse the WordPress dashboard while authenticated, the risk of full administrative session compromise is realistic and should be treated as a priority patch.

An attacker identifies a WordPress site running AI Engine 3.6.8 or earlier with a public-facing AI-powered input (such as a chatbot or content form) and crafts a URL or input payload embedding malicious JavaScript that the plugin reflects or stores without proper encoding. The attacker delivers this as a phishing link to a site administrator; when the administrator's browser renders the payload while authenticated, the injected script exfiltrates the WordPress session cookie or silently creates a rogue admin account. No confirmed public exploit code exists, but the attack class is well-understood and requires no specialized tooling.