In 90 days the WordPress security program received 2,004 vulnerability reports and paid out over $90,000 for them. That is almost half of everything it has paid since 2017. Until now one party covered the cost: Automattic. On Monday, October 5, 2026, Mary Hubbard, executive director of WordPress, wrote to the heads of more than 30 companies asking them to chip in.
I am basing this on The Repository article of October 8, 2026, which quotes Hubbard’s email (The Repository, October 8, 2026), and on the announcements on wordpress.org. I have not seen the email itself. I checked the figures and quotes against the sources on October 9, 2026.
What Mary Hubbard asked WordPress companies to fund
The email went out to the CEOs of hosting, plugin and security companies on Monday evening UTC. The Repository writes “Monday evening UTC” in its Thursday, October 8 piece. The main thesis: AI has made finding holes cheaper.
“AI had collapsed the cost of finding, reporting, and exploiting vulnerabilities”
Mary Hubbard, quoted by The Repository, October 8, 2026
According to her, the program received almost 2,000 reports in 90 days, more than ever, “and the pace is still climbing”. The Repository checked the program page on HackerOne: 2,004 reports and over $90,000 in bounties in 90 days, against a little over $200,000 paid out since the start in April 2017.
“Nearly half the program’s lifetime payouts happened in a single quarter, and the curve only goes one direction.”
Mary Hubbard, quoted by The Repository, October 8, 2026
Hubbard also writes that Automattic funded the program 100%: every bounty, triage, verification, fixes and coordinated security releases for all supported branches back to WordPress 4.7. She asks for one of three things: a contribution to a shared bounty pool, sponsoring people for triage and releases, or another arrangement worked out together. She stresses that part of the growth is real research that new tools make faster and that genuinely improves security. The problem, in her view, is the economics, not the program itself: “The economics are what’s broken, not the program”.
Which companies were asked to fund WordPress security
GoDaddy, Newfold Digital and Hostinger headed the list of recipients. The Repository goes on to name:
| Group | Companies |
|---|---|
| Hosts and registrars | SiteGround, IONOS, DigitalOcean, DreamHost, World Host Group, Kinsta, OVHcloud, group.one, Namecheap, Tucows, Pantheon, Porkbun |
| Plugins and products | Elementor, Awesome Motive, Incsub (WPMU DEV), Rocketgenius (Gravity Forms), Rank Math, Brainstorm Force, Extendify |
| Enterprise agencies | Human Made, rtCamp |
| Security | Wordfence, Patchstack, BlogVault |
Source: The Repository, October 8, 2026.
The email did not go to WP Engine, which has been in a legal dispute with Automattic since 2024, even though WP Engine representatives were thanked in the credits of the 7.0.2 and 7.1.1 security releases. Nexcess was not on the list either.
How hosts and plugin companies responded
Those who answered The Repository are in favor, but want to know exactly what they would be signing up for.
- Hostinger (Marco Chiesi): considering seconding its own people. Since July it has worked with the security team and deploys WAF rules before patches are published.
- Kinsta (Jon Penland): wants “a clearly defined program” that companies could contribute to.
- Awesome Motive (Syed Balkhi): already sponsors several contributors on the security and plugins teams, but notes that the request is “quite open-ended”.
- Human Made (Tom Willmot): sponsors John Blackbourn, who represents the security team full time. The agency previously tried to rally companies around a security fund, “unfortunately without much traction”.
- Patchstack (Oliver Sild): the company has paid out $600,000 in bounties from its own pocket since 2022, but was never admitted to the WordPress security and plugins teams. In his view “this process needs to be reformed”.
GoDaddy, Newfold Digital and Wordfence had not responded by the time the piece was published.
Is my WordPress site affected by the Apache HTTP/2 vulnerability?
Update, . On 11 November the Apache Software Foundation released Apache HTTP Server 2.4.71. The top entry on the Apache 2.4 vulnerabilities page is CVE-2026-119618, rated critical: a synchronisation flaw in mod_http2 (the assign_headers function) that can lead to remote code execution. It affects versions 2.4.55 to 2.4.70 and the standalone mod_h2 module from GitHub, 2.0.8 to 2.0.42. Apache credits Anthropic’s OSS Scanner (reported 26 October) and Google Cloud Vulnerability Research (3 November). AI scanners are now checking the web server your site runs on as well.

No WordPress, plugin or theme update fixes this. Apache handles the connection before a request ever reaches PHP. Wordfence’s PSA of 12 November points out that its firewall runs at the PHP layer, after Apache has processed the HTTP/2 connection, so the fix has to come from whoever runs the server. According to Wordfence, only servers with HTTP/2 enabled (h2 or h2c) are exposed, because Apache’s default protocol is HTTP/1.1. Wordfence also cites Red Hat (unauthenticated remote code execution when HTTP/2 is on) and Wiz (no known exploitation as of 11 November); I have not read those two sources myself. In the CVE record on NVD, the US agency CISA scores it 8.1 under CVSS 3.1 (high), with high attack complexity, and as of 13 November records no exploitation.
How to check whether your hosting is exposed
The WordPress dashboard shows your PHP version, not your Apache build, and not whether the HTTP/2 module is patched. The Server response header is often hidden or shows only the proxy in front. Behind Cloudflare you will see cloudflare there, and what matters is the origin server behind it. Wordfence notes that a CDN, nginx or a load balancer can serve HTTP/2 to visitors while Apache receives HTTP/1.1, so what your browser shows says nothing about the origin. With shell access you can check it like this:
# does the origin accept h2 over TLS? (2 = yes, 1.1 = not over TLS; the Protocols line decides h2c); ORIGIN_IP is the origin server's address
curl -sk --http2 --resolve example.co.uk:443:ORIGIN_IP -o /dev/null -w '%{http_version}\n' https://example.co.uk/
# is the module loaded, and where are protocols enabled?
apachectl -M | grep http2
grep -rn "Protocols" /etc/apache2/ /etc/httpd/ 2>/dev/nullThe version number alone does not settle it: Wordfence reminds readers that distributions and control panels backport fixes into older package numbers, so the operating system or panel vendor’s advisory is what counts.
On shared or managed hosting you will not have that access, so check your host’s status page or announcements first. If there is nothing there, open a support ticket:
“Hello, does the Apache server hosting my site (domain: …) accept h2 or h2c connections? Has the installed Apache package been patched for CVE-2026-119618, and was Apache restarted after the update? Please confirm in writing.”
If no fix is available yet, Wordfence recommends disabling HTTP/2 in Apache for now: remove h2 and h2c from the Protocols directive, virtual hosts included, restart, confirm the origin no longer accepts HTTP/2, and turn it back on once the patched package is installed. Server configuration belongs in a WordPress security audit as much as the plugin list does, and watching for host advisories like this one is part of WordPress website maintenance.
Which WooCommerce Stripe Gateway version should I install after the 11 November advisory?
Update, . The WooCommerce team published an advisory on 11 November for WooCommerce Stripe Gateway (Stripe for WooCommerce): update to 11.2.0, or to 11.1.2 if the store has to stay on the 11.1.x branch. The most significant of the fixed issues could have allowed fraudulent purchases; WooCommerce says it has no evidence they were exploited. A separate issue affected 10.9.0: with classic checkout, Express Checkout buttons, account creation at checkout and payment by Apple Pay or Google Pay, the customer’s new password was saved in plain text in the order data. From 11.1.0 the plugin removes those passwords and, if it removed any, emails the site administrator. No email does not mean you can skip the next check. If your store met those conditions, also check backups, order exports and any service your orders sync to. Versions 11.1.2 and 11.2.0 need WordPress 6.9 and WooCommerce 11.0. Check the version under Plugins > Installed Plugins even with automatic updates on, and do it on every copy of the shop, including a public staging site.
How many WordPress plugin vulnerabilities were disclosed in the first week of November 2026?
Update, . Wordfence’s weekly report for 2 to 8 November counts 497 vulnerabilities in 312 plugins and 1 theme, from 218 researchers. 457 are patched and 40 are not. By CVSS rating, 21 are critical, 118 high, 352 medium and 6 low. The most common types are cross-site scripting (179) and missing authorisation (108). For the 40 without a patch, updating changes nothing: the options are deactivating the plugin or replacing it. That is why your plugin list is worth comparing with a report like this every week, rather than waiting for a dashboard notice.
WordPress 7.1.4: 22 vulnerabilities and what they mean for your site
Update, . The two-month review of the Core Security Initiative shows how the team copes with the flood of reports. The queue of untriaged reports went from 359 to double digits, thanks to an AI-powered triage system that reads incoming reports, reproduces them securely and posts a summary and a recommendation (valid or not, how severe), while security researchers make the final call. Nearly 3,000 reports have been closed this year. The same tooling, which reads the code and checks whether a report holds up, is now pointed the other way: it hunts for vulnerabilities in core before anyone reports them. The team runs it with open weight models and purpose-built harnesses, and what it finds goes through the same release process. WordPress has also joined Anthropic’s OSS Scanner, which gives open source projects free scans with Anthropic’s strongest models; the first results are already being fixed, and Anthropic is credited in release announcements. On who pays: the post says the most useful help from organizations is sponsoring people for triage and release work and contributing to the bounty program.
Update, . WordPress 7.1.4 shipped on 28 October. The version page counts 22 vulnerabilities. The most serious is a race condition in the core update process that its reporter, Umut Bayram, says could lead to remote code execution. Most of the rest are stored XSS issues that need at least a Contributor account. Two reports are credited to Anthropic. The fixes were backported to older branches, from 7.0.8 down to 4.7.39, and sites with automatic background updates receive the patch on their own.
It is the seventh core security release since July. In the two-month update on the Core Security Initiative, Rudy Faile writes that the team has received more than 3,000 reports this year, more than the previous twelve years combined, and that 73% of the 109 steps in a security release now have automation. At that pace, leaving a security release until the weekend stops being a safe habit. In my WordPress update service security releases go to staging first, then to production.
Update, . The release note describes the top fix as “a weakness in the WordPress upgrade process which can lead to remote code execution during the upgrade window”. The core diff from 7.1.3 to 7.1.4 shows the mechanism in class-wp-upgrader.php and update-core.php. Until 7.1.3 the working directory under wp-content/upgrade/ was named after the package, so its path was predictable. 7.1.4 appends 16 random characters to that name, creates wp-content/upgrade/index.php if it is missing, empties the upgrade folder before it starts (keeping index.php) and no longer copies a wp-config.php from the working directory. Wordfence wrote on 29 October that its telemetry shows threat actors actively attempting to exploit the flaw. It deployed a firewall rule for its paid tiers on 28 October at 20:30 UTC, about three hours after the release; Wordfence Free gets the same rule on 27 November. Wordfence also advises checking Dashboard > Updates to confirm the site really runs 7.1.4 or the patched release of its branch, rather than assuming the background update completed. My advice for servers you control, and a point I check in every WordPress security audit: deny PHP execution under /wp-content/upgrade/ and look for leftover folders there after each update. In nginx:
location ~* ^/wp-content/upgrade/.*\.php$ { deny all; }With Apache 2.4, in wp-content/upgrade/.htaccess:
<FilesMatch "\.php$">
Require all denied
</FilesMatch>The index.php core writes there is only a silence file, and updates run from wp-admin rather than through HTTP requests into that folder, so the rule does not break updates.
What WordPress 7.1.3 fixes and who reported the bugs
The day after the email, on October 6, 2026, WordPress 7.1.3 shipped with 7 security fixes and 4 other bug fixes. The announcement credits the reporter for each fix:
| Vulnerability | Who can exploit it | Reporter |
|---|---|---|
| Stored XSS on the Comments screen in the dashboard, via pending comments | Anyone who leaves a comment, if a moderator (Editor or higher) clicks a link in it | Trail of Bits in collaboration with OpenAI |
Denial of service in WP_Http::make_absolute_url() | Contributor account or higher | Anthropic |
| Second-order SQL injection in the WXR export | An Administrator running an export, with a bad _thumbnail_id value already in the database | Anthropic |
| Author could make posts sticky | Author account or higher | Anthropic |
| Disclosure of comments on private and unpublished posts | Anyone, without logging in | Ananda Dhakal, Patchstack |
| XSS in Imgur embeds | Contributor account or higher and a victim who views the post | Zhengyu Liu, Jingcheng Yang, Gavin Zhong |
Forgeable {status}_{type} hook parameters | A plugin passing a raw status or post type to wp_insert_post() | Alex Concha, WordPress security team |
Sources: reporters per WordPress.org, October 6, 2026, exploitation conditions per Patchstack, October 6, 2026.
Only one of the seven holes works with no account and no victim involvement: the leak of comments from private posts. Patchstack explains the mechanism: in the comment feed of a single post, WordPress first fetched the comments and only then checked whether the visitor was allowed to see the post. The check hid the post but not the comments, and feeds do not return a 404, so the comments ended up in the feed. Three of the other six require a Contributor or Author account.
Four of the seven reports come from companies building AI models or their partners. That illustrates Hubbard’s thesis well: AI is now finding holes too, and someone has to pay for each of them with triage, a fix and a release for every supported branch. The announcement says the fixes are being backported to all branches covered by security patches, currently back to 4.7, and will be released as they become ready.
WordPress 7.1.1 and 7.1.2 security fixes
According to Patchstack, release 7.1.2 came out two weeks before 7.1.3 and fixed one vulnerability (CVE-2026-87902). It let an unauthenticated visitor trigger a local file inclusion and, with the right PHP configuration, led to remote code execution (Patchstack, October 6, 2026).
Earlier, in September, Patchstack also described a remote code execution vulnerability patched in 7.1.1. If your site is on 7.1.0, it is missing the fixes from three subsequent releases.
What is the WordPress Core Security Initiative
On August 28, 2026 the security team announced the Core Security Initiative on make.wordpress.org (Make WordPress Security, August 28, 2026). The announcement ties a “substantial increase” in reports to the rapid development of frontier AI models. It describes this as a good problem that nonetheless requires scaling triage, verification and fixing.
The initiative covers WordPress core. It does not cover plugins or themes. Reports are accepted through the HackerOne program, while vulnerabilities in WordPress.com and the mobile apps go separately, through the Automattic program.
How Patchstack and Wordfence filter AI-generated plugin vulnerability reports
The HackerOne figures cover core only. Plugin and theme holes go through other programs, and those show the same curve. In March 2026 the Wordfence Bug Bounty Program received 1,718 vulnerability submissions, 59.4% more than in February, from 321 active researchers, and paid $62,174 in bounties that month (Wordfence, March 2026 report). The report does not say how many of those submissions were written with AI help.
Patchstack names the cause directly. On May 29, 2026 Dave Jong, its security research lead, wrote that since AI-assisted research took off, “especially throughout 2026”, the program has been getting more lower-impact reports that only narrowly meet the criteria, more duplicates (one vulnerability drew more than 20 duplicate submissions from different researchers) and reports built on incorrect AI-generated assumptions (Patchstack, May 29, 2026). From June 1, 2026 Patchstack:
- removed the Contributor role from scope, so a hole has to be reachable by a guest, Subscriber, Customer or a similar low role,
- attached conditions to each vulnerability class, for example XSS has to have site-wide or reflected impact,
- limited the monthly bounty pool to the top 5 researchers,
- bans for one week anyone who submits incorrect AI-generated assumptions or a report not tested against the actual plugin or theme.
The guidelines as published on October 13, 2026 also require a step-by-step text proof of concept with raw HTTP requests and exact payloads, and a component with 1,000 or more active installs (Patchstack bug bounty guidelines).
For a plugin author this filter matters more than the core budget. A report that passes a platform’s triage still lands on the author, who has to reproduce it and ship a fix, usually without any bounty money of their own.
Update, . The flood reaches every open source registry. Node.js now accepts HackerOne reports only from researchers with a Signal of 1.0 or higher, after its security team received over 30 reports between 15 December and 15 January and said that triaging them took time away from legitimate security work (Node.js). The Node.js page speaks of low-quality reports. The AI link comes from the OpenJS Foundation security update for Q1 2026, which says the surge was “largely driven by AI-powered fuzzing and scanning tools” and reports that the Node.js bug bounty was paused after the Internet Bug Bounty ended its funding. Maintainers everywhere now ration the hours they spend on triage, and a plugin author with no bounty budget has fewer of them. That is why every WordPress maintenance contract I write sets aside hours for patching.
Who handles plugin vulnerabilities and CVEs on WordPress.org
The plugin directory has its own route. The Plugin Handbook asks reporters to contact the developer first, and only if that fails to email [email protected], without posting about the issue publicly. The plugins team then acts on the listing: “Most plugins are closed to prevent new downloads until the issue is resolved.” The same page says the team does not help with filing CVEs “due to a lack of resources” (Plugin Handbook, Reporting plugin security issues, last modified January 2, 2026).
So the CVE identifiers for WordPress plugins come largely from security companies registered as CVE Numbering Authorities. The CVE Program’s partner list gives Wordfence the scope of plugin, theme and core vulnerabilities discovered by or reported to its team, WPScan “WordPress core, plugins, and themes”, and Patchstack vulnerabilities found by its own team and its bug bounty participants (CVE Program, list of partners).
These companies pay for plugin triage out of their own revenue and sell protection on top of it. Wordfence writes that Premium, Care and Response users get firewall protection in real time, while free users get it after 30 days. That is the second answer to “who pays”: in plugins, the cost of triage already sits with security vendors, and two of them, Wordfence and Patchstack, are on Hubbard’s list.
What the EU Cyber Resilience Act requires from WordPress plugin vendors
Since September 11, 2026 part of this cost is also a legal duty in the EU. The Cyber Resilience Act, Regulation (EU) 2024/2847, applies in full from December 11, 2027, but Article 71(2) brought its Article 14 on reporting forward to September 11, 2026 (Regulation (EU) 2024/2847, EUR-Lex). A manufacturer that becomes aware of an actively exploited vulnerability in its product has to notify the designated CSIRT and ENISA through a single reporting platform:
| Step | Deadline under Article 14(2) |
|---|---|
| Early warning | 24 hours after becoming aware |
| Vulnerability notification | 72 hours after becoming aware |
| Final report | 14 days after a fix or mitigation is available |
Article 3(13) defines a manufacturer as someone who develops a product and markets it under its own name or trademark, “whether for payment, monetisation or free of charge”. A company that sells a premium plugin under its brand reads straight onto that wording. Article 13(6) adds that a manufacturer who finds a vulnerability in an integrated component, open source included, has to report it to whoever maintains that component and share the fix. Breaching Articles 13 and 14 can cost up to EUR 15 million or 2.5% of worldwide annual turnover, whichever is higher (Article 64(2)). Micro and small enterprises are spared fines only for missing the 24-hour early warning (Article 64(10)).
Open-source software stewards, legal persons that systematically support free software intended for commercial activities, get a lighter regime: a documented cybersecurity policy (Article 24) and no administrative fines at all (Article 64(10)(b)). The regulation does not say who pays for a steward’s triage, which is the same question Hubbard’s email puts to the companies on her list.
What WordPress site owners should do after 7.1.3
Hubbard’s email changes nothing on your site overnight. But the HackerOne numbers and the 7.1.3 list translate into a few concrete things:
- Install 7.1.3 if you have not yet. The announcement recommends updating immediately. Check under Dashboard, Updates that the site really is on 7.1.3, even if you have automatic updates enabled.
- Older branches get patches later. According to Patchstack, on release day the fixes reached branches from 6.6 upward, while branches 4.7 to 6.5 were still waiting. A site on 7.1 was protected from day one.
- Clear the cache after updating. Patchstack warns that 7.1.3 does not remove oEmbed embeds already stored in the database, so a malicious Imgur embed added earlier keeps displaying until you clear the cache.
- Review user roles. Patchstack recommends this explicitly, because three of the seven holes require a Contributor or Author account. Accounts of former authors and agencies that finished work long ago are worth removing or downgrading.
- Core is not everything. The Core Security Initiative does not cover plugins. If you have not reviewed your plugins in a long time, start with an audit of outdated plugins and known CVEs.
- There may be more security releases. Hubbard writes that the pace of reports is still climbing. If you update once a quarter, you can miss several releases in a row. With ongoing WordPress website maintenance, updates are tested on a copy and deployed within days, not months.
If you want to check the state of your site right now, start with a WordPress security audit.
When companies are expected to commit funding
Hubbard offered interested companies detailed program data and wrote that she wants a “first group of contributors standing with us this quarter”, meaning by the end of December 2026. So far no company has publicly announced a specific amount. I will add an update when one does.
Recommended update cadence after WordPress 7.1.3 and patching costs in maintenance agreements. Update 10.10.2026
As of October 10, 2026 I found no public pledge from any company on the list, so the funding picture above stands. The risk picture is easier to pin down. Patchstack rates 7.1.3 like this:
“Update as soon as you can, but this isn’t a drop-everything emergency.”
Patchstack, WordPress 7.1.3 security release, October 6, 2026
One of the seven fixes works with no account and no victim action, and Anthropic reported three. For a site owner that means a rhythm: core patch within days, after a test on a copy, not next quarter.
The money Hubbard asks for covers bounties and triage in core. The work on your own site (testing, deploying, checking plugins) is a separate cost, and the contract has to name it. Ask whoever handles your site maintenance three things: how many days after a security release they ship the core update, whether they test it on staging, and whether an out-of-cycle release is covered by the plan.
Last updated: October 13, 2026.







