An AI-Assisted Find in Nearly 18-Year-Old WordPress Code Reached CISA’s Exploited List Six Days After the Fix

A WordPress updater flaw found, its researcher says, with AI help was confirmed exploited six days after its fix, and five more flaws with reported exploitation or exploitation attempts need attention this week.

WordPress 7.1.4 shipped October 28 with fixes for 22 vulnerabilities. One of them, CVE-2026-119977, sits in the code that installs WordPress updates, and CISA added it to the Known Exploited Vulnerabilities catalogue on November 3, confirming active exploitation. That is six days from fix to confirmed attacks. CISA has not said who is exploiting it or how many sites are affected. Wordfence, which published the CVE record, scores the flaw 8.1 (High), while WordPress's own GitHub advisory rates it low (CVE record, advisory). Published advisories list versions 4.7 through 7.1.3 as affected; Picus traces the behaviour back to WordPress 2.7, released in 2008. The fix is backported to branches down to 4.7. Sources: WordPress 7.1.4, CISA.

Umut Bayram of Picus Security, who reported the flaw, says he used GPT 5.6 Sol throughout the work and puts the effort at about four hours, from first reading the update process to a working exploit chain. The four-hour figure is Bayram's own account, and nothing published ties the November exploitation to AI tools. The business point is the timeline: roughly four hours of model-assisted work against behaviour that dates to 2008 produced a working exploit, and your patch policy runs against that clock. Source: Picus Security.

Start with a test your hosting team can run today: request the staging folder, wp-content/upgrade/, from the public internet. WordPress unpacks update files there, including a temporary installer, before copying them over the live site. In affected versions the copy step does not skip wp-config.php, so a configuration that an attacker had the installer write, pointing at a database they control, replaces the live one. That needs the server to answer and run PHP in the folder, and the attacker has to win a timing race. Picus describes the follow-on: an admin account appears and a plugin upload runs code. The attacker also has to know the staging path, which is predictable on current branches (5.9.0 through 7.1.3). Sources: Picus, Hackread.

Updating to 7.1.4 runs through the same updater, and WP-CLI's core update command uses the same path, so the first update still carries the exposure. Before updating, block the staging folder at the web server, then confirm that a request to it is refused. Releases from 2.7 through 4.6 received no fix and need a full upgrade. One more data point from the same release: Anthropic is credited with reporting two of the 22 fixes, a stored cross-site scripting issue and a hardening fix. An AI lab sits on the credit list beside independent researchers. Sources: WordPress 7.1.4, Picus. Ask your team this week which of your WordPress sites, including agency-hosted ones, can be reached at the staging folder today.

The Cyber Centre covered WordPress and SonicWall. Two recent privacy files show what a large breach costs a Canadian organization afterwards.

The Canadian Centre for Cyber Security issued advisory AV26-1117 on November 3, covering CVE-2026-119977 in WordPress versions before 7.1.4. For SonicWall, it issued AV26-1133 on November 6, naming CVE-2026-19868 and citing SonicWall's indication that it is being exploited. Canadian organizations therefore have a domestic source to cite in change records and board updates. Sources: CCCS AV26-1117, CCCS AV26-1133.

What a breach costs after the incident is visible in the Office of the Privacy Commissioner's October 26 announcement on Loblaw. An intrusion in February 2026 exposed customer account information for about two million Canadians. Loblaw's October 21 compliance letter commits it to an outside assessment of its upgraded defences, a report to the OPC on how it will act on that assessment, and wider security improvements. The file closes only when the Commissioner judges the promised work complete. The release refers to Canada's federal private-sector privacy law. A director should plan on an independent assessment and ongoing regulator reporting as the likely price of a large breach. Source: OPC.

The Burnaby School District shows the same cost curve at a smaller organization. A cyberattack on September 21 led, by the district's October account, to theft of personal information about students, parents and staff, and the district notified the RCMP and B.C.'s Office of the Information and Privacy Commissioner. It says residency documents and some students' support-level details may be among the records taken. Source: CP24.

This period's active exploitations and notable disclosures, mapped to the control that contains them.

Threat: An account that can also sign in by another method skips the SAML step that enforces MFA on the VPN. SonicWall rates it 7.1, says it looked into a report of possible exploitation, and lists no workaround. The Cyber Centre issued AV26-1133 the same day. SecurityWeek, CVE record, CISA

Defence: Upgrade to SonicOS 7.3.4-7013 (Gen7) or 8.2.3-8015 (Gen8), then review VPN logins. Look for SAML users with a successful sign-in by another method, and ask who can still sign in outside SAML. A bypassed login looks routine in the logs.

Threat: Volexity reports the two flaws were chained with the earlier bypass CVE-2026-104935 against fully patched devices, with exploitation since at least October 3. Ivanti updated its advisory with fixes. CISA, BleepingComputer

Defence: Apply the updated advisory's fixes; assume compromise if exposed earlier. Ask who owns the appliance, whether it was patched against the updated advisory, and whether sessions and credentials were rotated.

Threat: The Node-RED web interface needs no login, so a remote attacker can run commands. VulnCheck reported exploitation attempts against its honeypots from October 30, including a backdoor that survives reboots. Siemens fixed it in August (SSA-834709, Industrial OS V4.3.4-1). The Hacker News

Defence: Update to V4.3.4-1 or remove Node-RED; do not expose the interface. A patch released in August sitting unapplied in November is a governance finding. Ask operations technology owners for gateway versions and whether any reach the internet.

Threat: Sekoia attributes watering-hole attacks on iPhones and Android phones to APT29 with moderate-to-high confidence. The chains use patched flaws to steal browser cookies, and Sekoia says fully updated devices are not affected. Sekoia, The Hacker News, CISA

Defence: Enforce minimum OS versions on phones that reach corporate data. Ask what share of managed and personal devices run current iOS and Android patches, and what happens to those that do not.

Threat: Affects OpenSSH on Linux systems built on musl (Alpine's default) with PAM disabled, where an account with no password set (marked with an asterisk) can be entered using a password of 256 to 1,024 characters. Previdian confirmed exploitation attempts from November 6, and its founder told Help Net Security an attacker gained root on a Previdian Alpine honeypot. Previdian, Help Net Security

Defence: Upgrade to OpenSSH 10.6p2 or your distribution's patched package, then restart sshd. Glibc distributions and systems using PAM are unaffected. Ask whether Alpine-based images or appliances expose SSH to the internet.

HARDENED is published for general informational and educational purposes. All threat data is sourced from publicly available security research and cited accordingly. This newsletter does not constitute professional security advice. Security configurations and threat landscapes vary by organization. Consult a qualified security professional for implementation guidance specific to your environment. All data as of November 8, 2026.