CVE-2026-89093: WordPress Better Messages Auth Bypass Flaw

CVE-2026-89093 Overview

CVE-2026-89093 affects the Better Messages – Chat Rooms, Group Chat, Private Messages & AI Chat Bots plugin for WordPress in all versions up to and including 2.15.33. The vulnerability enables unauthenticated attackers to impersonate the plugin's internal AI bot accounts by spoofing the X-Real-IP HTTP header during guest registration. Once the plugin misidentifies the attacker as a trusted AI bot, per-room role allowlists, draft-status checks, and join filters are bypassed. Attackers can then enter administrator-restricted chat rooms, post messages, and read private message history belonging to other users. The weakness maps to [CWE-287: Improper Authentication].

Critical Impact
Unauthenticated attackers can read private messages between other WordPress users and inject content into restricted chat rooms by spoofing a single HTTP header.

Affected Products

Better Messages – Chat Rooms, Group Chat, Private Messages & AI Chat Bots plugin for WordPress
All versions up to and including 2.15.33
Fixed in version 3.0.0

Discovery Timeline

2026-09-19 - CVE-2026-89093 published to the National Vulnerability Database
2026-09-21 - Last updated in NVD database

Technical Details for CVE-2026-89093

Vulnerability Analysis

The flaw resides in how the plugin determines whether a chat participant is a privileged internal AI bot. The is_ai_bot_user() function in inc/functions.php performs a prefix match for the string 'ai-chat-bot-' against a guest record's stored IP address field. That field is populated verbatim from the client-controlled X-Real-IP request header when an unauthenticated visitor registers as a guest through the routines in inc/guests.php.

Because the bot check short-circuits both user_can_join() and user_can_read() in inc/chats.php, a positive match bypasses the per-room role allowlist, the draft-status check, and any join filters. The attacker gains the same room-access surface as a trusted internal automation account without ever authenticating.

Exploitation impacts confidentiality: the private message history of other users, including content posted in administrator-restricted rooms, becomes readable. See the Wordfence Vulnerability Analysis for additional context.

Root Cause

The root cause is a trust boundary violation. A security-relevant identity decision (is this actor an internal AI bot?) is made against attacker-controlled data (the X-Real-IP header stored as the guest IP). The plugin never validates that the IP field contains a real IP address, nor does it separate bot identity from guest storage. Any string beginning with ai-chat-bot- submitted through the header is accepted as proof of bot identity.

Attack Vector

The attack is remote, unauthenticated, and requires no user interaction. An attacker sends a guest-registration request to the plugin's endpoint while supplying an X-Real-IP header whose value begins with ai-chat-bot-. The resulting guest record is treated as an internal bot on every subsequent request. The attacker can then call chat read and post endpoints against rooms they should not be able to enter. Relevant source locations include chats.php line 2145, chats.php line 2196, functions.php line 4155, and guests.php line 311.

No verified public proof-of-concept code is available at time of writing. Detailed technical breakdowns are published by Wordfence and in the plugin source references above.

Detection Methods for CVE-2026-89093

Indicators of Compromise

Guest registration requests containing an X-Real-IP header value that starts with the string ai-chat-bot-.
Database rows in the Better Messages guest table where the stored IP column contains non-IP text, especially values prefixed with ai-chat-bot-.
Chat room join or read activity attributed to guest accounts in administrator-restricted rooms.

Detection Strategies

Review web server and reverse proxy logs for inbound X-Real-IP header values that are not valid IPv4 or IPv6 addresses.
Query the plugin's guest storage for records whose IP field fails a strict IP-format regex.
Correlate guest registration events with immediate access to private or restricted rooms by the same session identifier.

Monitoring Recommendations

Enable verbose logging on the WordPress REST and AJAX endpoints exposed by bp-better-messages.
Alert on any single client that registers a guest and then reads messages across multiple rooms in a short interval.
Monitor for outbound scraping patterns where guest sessions pull full message history rather than participating conversationally.

How to Mitigate CVE-2026-89093

Immediate Actions Required

Upgrade the Better Messages plugin to version 3.0.0 or later on every affected WordPress site.
Audit existing guest records and delete any entries whose stored IP field contains non-IP data.
Review recent chat access logs for unauthorized reads of restricted rooms and notify affected users if private message history was exposed.

Patch Information

The vendor addressed the issue in version 3.0.0. The code changes are visible in the WordPress Function Changeset between tags 2.15.33 and 3.0.0. Administrators should apply the update through the WordPress plugin manager and confirm the installed version reports 3.0.0 or higher.

Workarounds

Configure the reverse proxy or web application firewall to strip or overwrite inbound X-Real-IP headers on requests to the WordPress site.
Block requests to guest registration endpoints where the X-Real-IP header does not parse as a valid IP address.
Temporarily disable the Better Messages plugin on sites that cannot be patched immediately, especially those hosting sensitive private conversations.

bash

# Example nginx snippet: overwrite client-supplied X-Real-IP with the real connection IP
location / {
    proxy_set_header X-Real-IP $remote_addr;
    proxy_set_header X-Forwarded-For $remote_addr;
    proxy_pass http://wordpress_upstream;
}