On October 21, PHP, the programming language behind a large share of the web, patched two flaws that let an attacker take over a web server without logging in. The researchers who found them, with help from AI models, named the pair PHP2Shell, and most of the coverage is about the AI. I am more interested in what the episode does to “reasonable security,” the legal standard that decides who pays after a breach. That standard rests on two assumptions: that you have weeks to install a patch, and that a weakness nobody has managed to exploit can be written down and left alone. PHP2Shell puts pressure on both. I wrote this on the day the patch came out, while the facts were changing by the hour (is anyone actually attacking servers yet? nobody could say for sure). Where it matters, I say what was known at the time.
The PHP project released versions 8.5.12, 8.4.27, 8.3.36 and 8.2.35 to fix two vulnerabilities, CVE-2026-109618 and CVE-2026-109619, each rated 9.2 under CVSS v4.0. Each can let an unauthenticated attacker run code on a web server with the privileges of the PHP worker. The flaws sit in the runtime itself, so a site can be exposed without any bug in its own code. The project’s advisory lists every common way of serving PHP to the web as affected (PHP-FPM, Apache with mod_php, php-cgi, FrankenPHP and LiteSpeed) and rates the risk highest for mod_php, where “an unauthenticated request can trigger the attack before any application code runs.” PHP’s advisories skip the branding and stick to plainer titles (“RCE via OOM corruption” and “RCE via stale global state”).[1] The Wall Street Journal’s headline puts “up to 70%” of the world’s websites at risk. The closest public number is W3Techs’ count of PHP on 69.3 percent of websites whose server-side language it can identify, which measures market share rather than exposure.[2]
Calif, a security firm auditing PHP under OpenAI’s Patch the Planet program, found its first trigger on August 1, a few days into the audit. It had a working remote code execution proof of concept in PHP’s hands by August 14. The Journal reports that the model involved was OpenAI’s GPT-5.6 Sol. Calif’s own write-up stresses that the initial discovery used no special harness and no model that was not generally available.[3] Wiz found variants independently and reported them on August 24, using its own tooling alongside Google and OpenAI models.[4] PHP’s two advisories credit four separate reporters. The embargo held for ten weeks and broke on October 20, when a Hacker News post announced two critical PHP remote code execution bugs set for disclosure on the 22nd. PHP published an advance notice that evening and moved the release up a day.[5]
Both assumptions behind “reasonable security” are really estimates of what it costs an attacker to build a working exploit. PHP2Shell is a public, well-documented instance of that cost falling, and the standard of care moves with it.
A Bug Class with a Paper Trail
Calif’s write-up includes a short history of the bug class. In 2004 Stefan Esser found what Calif calls “the first PHP memory limit remote vulnerability,” serious enough that he kept his exploit private. In 2020 a bug report (#78876) flagged the out-of-memory behavior again, and according to Calif, “not much was made of the report.” In 2024 the audit of PHP’s core that OSTIF commissioned from Quarkslab turned up a related out-of-memory issue. Again per Calif, “exploitation was not proven, hence the bug was not fixed.”[6]
The mechanism is old-fashioned C. When an allocation would push a request past PHP’s memory limit, PHP has long bailed out with longjmp(), abandoning whatever it was doing and unwinding to a safe point. If that happens while an internal structure is half-built, request cleanup can later treat the half-built structure as valid. A related path leaves state that a worker keeps between requests pointing into memory that has already been freed, for the next request to pick up. Calif’s route to memory exhaustion runs through multipart form parsing, which PHP performs before the application’s own code runs. Put plainly, if PHP runs out of memory at the wrong moment, it leaves debris behind and then trusts the debris. The fix removes the longjmp() path for out-of-memory errors and replaces it with a delayed soft limit and a hard exit.[7]
Both advisories still score attack complexity as High. In CVSS v4.0, that means a working exploit depends on getting around protections the platform already has in place, such as memory layout randomization.[8] That remains true of PHP2Shell. It is a smaller obstacle than it used to be when a researcher who knows the target can get from first trigger to working exploit in two weeks with a model’s help.
None of that history looks unreasonable in real time. Declining to fix a finding that nobody has shown to be exploitable is how security teams with finite budgets triage. As Calif tells it, the 2024 decision turned on the missing exploit. Underneath a decision like that is a judgment about cost: proving exploitability takes expensive expert time, and until someone spends it, the finding stays theoretical. That judgment was defensible when it was made, but it came with no expiration date.
Most companies keep documents like this. A risk register entry accepts a known weakness because exploitation is theoretical. A scanner exception is approved because the vulnerable function is unreachable. A memo keeps an end-of-life runtime in production because the upgrade costs more than the risk. Each records a judgment about attacker cost, made on the evidence available at the time. Each also proves that the company knew about the weakness, which a plaintiff or an examiner would otherwise have to establish circumstantially. When the Equifax multidistrict litigation reached the motion to dismiss, the court found a duty of care under Georgia law because, on the allegations, Equifax “knew of a foreseeable risk to its data security systems.” The same opinion recounts that Equifax “internally disseminated the warning, but never implemented the patch.”[9]
Companies will keep accepting risk, and they should. The change is to treat acceptances that rest on exploit difficulty as perishable. They need a review date and a trigger that reopens it when the underlying assumption moves. A public proof of concept is the obvious trigger. A credible report that a bug class has gone from theoretical to practical with AI assistance should now be another, and PHP2Shell is that report for memory-exhaustion bugs in PHP.
The Window
The other assumption is about time. The card industry’s security standard still gives merchants one month to install patches for critical vulnerabilities.[10] The federal government has already moved. On June 10, CISA revoked the directive that had given federal civilian agencies two weeks to fix a vulnerability once it appeared in the Known Exploited Vulnerabilities catalog. Its replacement, BOD 26-04, sorts vulnerabilities by public exposure, catalog status, automatability and impact. A catalog entry that gives an attacker total control of a system now gets three days, along with a forensic check for existing compromise. Most other catalog entries keep fourteen. CISA pointed to adversaries’ use of AI as one reason the window needed to shrink.[11] Mandiant’s data shows the trend predates the models. Its average time-to-exploit fell from 63 days in 2018 and 2019 to 32 days in 2021 and 2022, and then to five days in 2023, a figure that excludes fifteen outliers (with them, the average is 47).[12]
PHP2Shell shortens the interval again. Cloudflare had protective rules on by default for its free and paid WAF customers from 03:02 UTC on October 21, hours before the release. It still reported attack traffic against its customers, with attackers “mutating the attack payload” to get past the rules.[13] Check Point’s analysis states the mechanism directly: once the patch is public, “anyone can diff it to build a working exploit.”[14] Thai Duong, Calif’s chief executive, put it to the Journal less formally: “Some dude on Twitter could take the patches and run AI against it and post some exploit script.”[15] Wiz says its own team could “develop and adapt working exploits across multiple applications in hours,” and credits frontier models for the speed.[16]
Government guidance moved just as quickly. The UK’s National Cyber Security Centre told organizations to find every PHP system and patch it, and, if mitigations could not be applied in time, to “consider taking the system offline until it can be fully mitigated.”[17] CERT-EU, which serves the EU institutions, advised that where immediate remediation was not feasible, organizations should consider “isolating internet-facing applications from public networks until a fix can be applied.”[18] Valtori, the Finnish government’s IT services center, shut some public services down as a precaution.[19] As of this writing, CISA had not added either CVE to its catalog, and nobody had publicly confirmed a successful compromise.[20]
The T.J. Hooper Problem
Tort law has seen this before. In March 1928 two tugs lost the coal barges they were towing in a gale off the New Jersey coast. The tugs had no working radio receivers (the masters’ own sets were out of order), so they missed the storm warnings, and there was no general custom on the coast of carrying them. The owner’s defense came down to having done what the industry did. Learned Hand rejected it: “a whole calling may have unduly lagged in the adoption of new and available devices. It never may set its own tests, however persuasive be its usages.”[21]
A month to install a critical patch is a usage of exactly that kind. Following it is still evidence of reasonable care, and in most cases persuasive evidence. It is weaker evidence than it was last week for an unauthenticated remote code execution bug in an internet-facing runtime with attack traffic on the first day. The vendor told users to update “immediately.” Two cybersecurity agencies told anyone who could not to consider isolating the system or taking it offline.[22] And CISA has already decided, for its own agencies, that the worst catalog entries get three days.
Hand’s later formula in United States v. Carroll Towing Co. supplies the arithmetic. A precaution is required when its burden (B) is less than the probability of the harm (P) multiplied by the gravity of the harm (L).[23] PHP2Shell does not change what a breach costs. It changes P, and it changes how fast P climbs after disclosure.
A worked example, with illustrative numbers. Take an online retailer running its storefront on PHP across a few dozen servers, with no web application firewall in front. Patching today instead of in the monthly maintenance window costs about $30,000 in overtime, compressed testing and a few hours of degraded checkout. A serious compromise of the storefront would cost about $3 million in forensics, notification, card-brand assessments and lost sales.
Suppose that under the old timeline, the chance of this retailer being compromised through this bug before the monthly window is 0.5 percent. PL is $15,000. That is less than B, and waiting for the window passes the test.
Now suppose the chance is 5 percent, because the bug is unauthenticated and public, attack traffic is already visible, and anyone who can diff a patch with a model’s help can build an exploit. PL is $150,000, five times B. The retailer, its servers and the patch are unchanged; only P has moved.
The example also shows where the cheaper precautions were. Cloudflare’s rules were on before the patch was public, and PHP published interim configuration changes with the release. Neither is a complete fix. Cloudflare cautions that request data alone “does not always provide enough evidence to distinguish an exploit from legitimate traffic,” and PHP says its interim settings “do not make an installation safe.”[24] But an interim control, an isolation decision and a shutdown are all ways of acting the same day. The new federal directive builds this in, since removing a system’s public exposure moves it into a slower tier.[25] The duty this points to is a duty to respond within hours, not a duty to finish patching within hours, and testing can still take the time it takes.
Where the Duty Bites
A negligence suit is the cleanest frame for this argument and probably not the first place it shows up. Data breach plaintiffs still have to show a concrete harm to establish standing under TransUnion LLC v. Ramirez, and the economic loss rule ends many negligence claims, depending on the state.[26] The Equifax court found an independent duty that kept the consumer negligence claims clear of Georgia’s economic loss rule. That was 2019, two years before TransUnion, on a record where Equifax did not press a standing challenge.[27] Not every court will get there. The pressure is more likely to arrive through regulators, contracts and insurers first.
The Federal Trade Commission has been explicit for years. Responding to Log4j in January 2022, it said it “intends to use its full legal authority to pursue companies that fail to take reasonable steps to protect consumer data from exposure as a result of Log4j, or similar known vulnerabilities in the future.” It pointed to Equifax, where “a failure to patch a known vulnerability irreversibly exposed the personal information of 147 million consumers.”[28] The position survived the change in leadership. Its case against GoDaddy, resolved by an order finalized 3-0 in May 2025, alleged that the company failed to “inventory and manage assets and software updates” across its shared hosting services. The order goes further on end-of-life software. GoDaddy must disconnect its hosting environment from hardware running unsupported GoDaddy-managed software or, where that is infeasible, put temporary controls in place and document a plan to disconnect.[29] Shared hosting is where a great deal of the internet’s PHP runs.
For regulated crypto firms the standard is already in a rule. New York’s cybersecurity regulation requires covered entities, virtual currency licensees among them, to keep a monitoring process that surfaces new vulnerabilities promptly and to “timely remediate vulnerabilities, giving priority to vulnerabilities based on the risk they pose to the covered entity.”[30] The Department of Financial Services has enforced Part 500 against a BitLicensee before. Robinhood Crypto’s $30 million consent order in 2022 covered cybersecurity violations alongside anti-money laundering ones.[31] “Timely” carries no number. An examiner reviewing how a firm handled PHP2Shell is likely to measure it against what was knowable on October 21, including the vendor’s “update immediately” and the NCSC’s advice to consider taking unpatched systems offline.
Commercial contracts reach that result by another route. Security covenants in master services agreements and data processing agreements usually promise measures “consistent with industry standards” or “commercially reasonable” ones. Those phrases take their content from what the industry actually does. This week that included edge rules deployed before the patch was public and government services taken offline as a precaution. A vendor that held PHP2Shell for its monthly window will be defending that choice against that record.
Insurers already have the data. Coalition’s 2023 claims report found that policyholders with even one unresolved critical vulnerability were “33% more likely to experience a claim,” and that policyholders running end-of-life software were three times more likely to have an incident.[32] Underwriting questionnaires ask about patch cadence, and the answers are representations in the application.
Then there is Europe. Since September 11, Article 14 of the Cyber Resilience Act has required manufacturers of products with digital elements to notify “any actively exploited vulnerability contained in the product” to their coordinating CSIRT and to ENISA. The early warning is due within 24 hours of becoming aware of it, a fuller notification within 72 hours, and a final report no later than 14 days after a corrective or mitigating measure is available. Article 69(3) extends the obligation to products placed on the market before December 2027, so it covers the installed base.[33] Plenty of commercial products ship PHP inside them, often behind a web administration console. The NCSC’s guidance flagged exactly this, telling organizations to watch for updates from “third-party software vendors” whose products bundle PHP and pointing to the Dutch NCSC’s inventory of products with a PHP dependency.[34]
The trigger is a definition. An “actively exploited vulnerability” is one “for which there is reliable evidence that a malicious actor has exploited it in a system without permission of the system owner.” Attack traffic is not that. A confirmed compromise is. On October 21 the public record sat between the two, which put every manufacturer with PHP in an EU product through that analysis at once, against a clock that runs from awareness. Last month Edward Machin at Ropes & Gray asked whether a vulnerability that cannot realistically be exploited in a manufacturer’s own implementation still triggers Article 14. He concluded that reading the Act by reference to the product as implemented is “a useful product-specific analytical framework, but not a clearly established reporting exemption.”[35] PHP2Shell adds a second caution. “Not exploitable in our product” is a statement about attacker cost on the day it was written.
The Crypto Version
Crypto teams tend to spend more on auditing smart contracts than on the servers that deliver the app. The contracts hold the assets. The front end tells users’ wallets what to sign. BadgerDAO is the reference case. In December 2021 an attacker used a compromised Cloudflare API key to inject a script into the protocol’s front end. The script added unlimited token approvals for the attacker’s address to users’ transactions, about 500 wallets signed them, and roughly $120 million left, although the contracts worked as written throughout.[36]
PHP2Shell offers a server-side route to that result. Code execution on the server that delivers a dapp’s JavaScript, on the WordPress marketing site that links to the app, or on the docs portal users trust is enough to put a drainer in front of them without touching a contract. Sansec’s guidance to Magento merchants (block multipart requests at the web server, go read-only or offline) is the card-skimming version of the problem.[37] For a protocol that describes itself as decentralized, the front end is usually the component with an identifiable operator, which makes that operator the natural defendant. For a BitLicensee, all of it is Part 500 work, with a 72-hour notification clock if it becomes a cybersecurity incident.[38] Everyone else carries the exposure without the rulebook.
The Order of Operations
What I would run this week, in order:
Inventory every PHP runtime, including the ones you did not install. Marketing sites, docs portals, admin panels, staging servers and products that bundle PHP all count. For anything hosted, get the provider’s patch status and timing in writing.
Patch, or put an interim control in place the same day, and record which. The fixed releases are 8.5.12, 8.4.27, 8.3.36 and 8.2.35, and Linux distributions received the patches in advance. PHP’s interim settings, a WAF rule, isolation and a shutdown are all legitimate stopgaps, but none of them is the fix.
Look for compromise before declaring victory. The NCSC asks organizations that cannot update or disable a system immediately to check it for evidence of compromise first, and CISA’s new directive pairs its three-day deadline with a forensic check for that reason. A patched server that was already compromised is still compromised.
Write down what you did and when. Reasonableness will be judged later against a timeline of public guidance that is already fixed. A contemporaneous log of decisions and timestamps is the best evidence an operator will have.
Run the notification analysis in parallel. Part 500’s 72-hour clock, the Cyber Resilience Act’s 24-hour early warning and the notice provisions in customer contracts run from awareness or determination. Patching does not stop any of those clocks.
Reopen the risk acceptances that assumed exploitation was hard, and the ones that kept an end-of-life runtime alive. PHP 8.1 and older get no release from the PHP project, only source backports for 7.4, 8.0 and 8.1 (some LTS vendors have their own). W3Techs counts about a third of PHP sites on PHP 7 or PHP 5.[39] Give every exploitability-based acceptance a review date.
Rewrite the patch policy before the next one. Add an emergency tier for unauthenticated remote code execution in internet-facing components, measured in hours and satisfied by either the patch or a documented compensating control. Keep the one-month window for everything else.
Wiz’s co-founder told the Journal that PHP2Shell is “the most severe AI vulnerability that was ever found.”[40] That superlative will not last long. The cost of turning a known weakness into a working exploit has dropped, and the standard of care was always a function of that cost. For its own systems, CISA now counts in days. The card industry’s standard still says a month.
Notes
[1] PHP Group, Security Advisory: Critical Remote Code Execution Vulnerabilities, Update Now (Oct. 21, 2026), https://www.php.net/archive/2026.php#2026-10-21-5 [hereinafter PHP Security Advisory]; php-src Security Advisory GHSA-w8hw-w9fx-j79w, RCE via OOM corruption (CVE-2026-109618) (Oct. 21, 2026), https://github.com/php/php-src/security/advisories/GHSA-w8hw-w9fx-j79w; php-src Security Advisory GHSA-6wmf-4rv5-p3rp, RCE via stale global state (CVE-2026-109619) (Oct. 21, 2026), https://github.com/php/php-src/security/advisories/GHSA-6wmf-4rv5-p3rp.
[2] Robert McMillan, AI Just Found a Bug That Puts up to 70% of the World’s Websites at Risk, Wall St. J. (Oct. 21, 2026), https://www.wsj.com/tech/cybersecurity/ai-just-found-a-bug-that-puts-up-to-70-of-the-worlds-websites-at-risk-fd0902ea; W3Techs, Usage Statistics and Market Share of PHP for Websites (data as of Oct. 21, 2026), https://w3techs.com/technologies/details/pl-php.
[3] Calif, PHP2Shell, https://calif.io/php2shell/ (timeline: audit begun July 30, 2026; first trigger Aug. 1; first report to PHP Aug. 11; proof of concept shared Aug. 14); McMillan, supra.
[4] Wiz Research, PHP2Shell: AI Uncovers Critical RCE Vulnerabilities Affecting 70% of the Internet (Oct. 21, 2026), https://www.wiz.io/blog/php2shell-php-rce (naming Wiz Atlas, Gemini 4 Argon and GPT-5.6 Sol); McMillan, supra (Wiz’s exploits were “discovered independent of Calif”).
[5] GHSA-w8hw-w9fx-j79w, supra, and GHSA-6wmf-4rv5-p3rp, supra (credits); Check Point, php2shell: What We Know About the Leaked Critical PHP Vulnerabilities (CVE-2026-109618 & CVE-2026-109619) (Oct. 21, 2026), https://blog.checkpoint.com/security/php2shell-what-we-know-about-the-leaked-critical-php-vulnerabilities-cve-2026-109618-cve-2026-109619 (advance notice published 19:33 UTC on Oct. 20; release set for 14:00 UTC on Oct. 21, one day before the embargo was to end).
[6] Calif, supra. The audit’s public summary lists 27 findings, two of them redacted because their fixes were “complex and in progress”; it does not describe an out-of-memory finding. Quarkslab, Security Audit of PHP-SRC (Apr. 10, 2025), https://blog.quarkslab.com/security-audit-of-php-src.html.
[7] GHSA-w8hw-w9fx-j79w, supra (“PHP uses longjmp() for out-of-memory (OOM) errors to exit the current context and abort handling of the request.”); GHSA-6wmf-4rv5-p3rp, supra (“A fatal error, including one caused by memory exhaustion, can abandon an operation before it performs its matching reset.”); Calif, supra.
[8] Both advisories carry the vector CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N. See FIRST, Common Vulnerability Scoring System Version 4.0: Specification Document (Attack Complexity metric), https://www.first.org/cvss/v4.0/specification-document.
[9] In re Equifax, Inc., Customer Data Sec. Breach Litig., 362 F. Supp. 3d 1295, 1310, 1325 (N.D. Ga. 2019).
[10] PCI Security Standards Council, Payment Card Industry Data Security Standard v4.0.1, Req. 6.3.3 (June 2024) (requiring patches for critical vulnerabilities to be installed within one month of release).
[11] Cybersecurity & Infrastructure Sec. Agency, Binding Operational Directive 26-04, Prioritizing Security Updates Based on Risk (June 10, 2026), https://www.cisa.gov/news-events/directives/bod-26-04-prioritizing-security-updates-based-risk (revoking BOD 22-01; full timelines apply within 180 days); Tenable, CISA BOD 26-04: Frequently Asked Questions About the New Risk-Based Patching Directive, https://www.tenable.com/blog/cisa-bod-26-04-FAQ-vulnerability-remediation-impact; Cloud Security Alliance, CISA’s BOD 26-04: A Risk-Based Reset of Patch Rules, https://labs.cloudsecurityalliance.org/research/csa-research-note-cisa-bod-26-04-risk-based-vulnerability-re/; CISA BOD 26-04 Directs Agencies to Prioritize Exploited Vulnerabilities and Assess Compromise Before Patching, Industrial Cyber (June 11, 2026), https://industrialcyber.co/cisa/cisa-bod-26-04-directs-agencies-to-prioritize-exploited-vulnerabilities-and-assess-compromise-before-patching/.
[12] Casey Charrier & Robert Weiner, How Low Can You Go? An Analysis of 2023 Time-to-Exploit Trends, Google Cloud Blog (Oct. 15, 2024), https://cloud.google.com/blog/topics/threat-intelligence/time-to-exploit-trends-2023.
[13] Cloudflare, Cloudflare Deploys Protections Against php2shell Vulnerability (Oct. 21, 2026), https://blog.cloudflare.com/php-vulnerabilities-waf-protection/.
[14] Check Point, supra. Check Point cites CVE-2024-4577, an earlier PHP flaw that drew botnet scanning within about a day of disclosure.
[15] McMillan, supra.
[16] Wiz Research, supra.
[17] Nat’l Cyber Sec. Ctr. (U.K.), Critical Security Issue Affecting PHP (Oct. 21, 2026), https://www.ncsc.gov.uk/news/critical-security-issue-affecting-php.
[18] Update Issued for Critical PHP Vulnerabilities: Security Agencies Warn Patch Now or Take Systems Offline, ITPro (Oct. 21, 2026), https://www.itpro.com/security/fixes-issued-for-critical-php-vulnerabilities-security-agencies-warn-patch-now-or-take-systems-offline (quoting CERT-EU).
[19] Precursor Security, Live Blog: PHP2Shell Critical PHP RCE Vulnerabilities (CVE-2026-109618 & CVE-2026-109619) (Oct. 21, 2026), https://www.precursorsecurity.com/blog/php2shell-critical-php-rce-vulnerabilities (Valtori stated the shutdowns were not a response to a breach).
[20] PHP Issues Emergency Patches for Two Critical PHP2Shell RCE Flaws as Exploit Attempts Emerge, The Hacker News (Oct. 21, 2026), https://thehackernews.com/2026/10/php-issues-emergency-patches-for-two.html; Check Point, supra.
[21] The T.J. Hooper, 60 F.2d 737, 740 (2d Cir. 1932).
[22] PHP Security Advisory, supra (“All PHP users are strongly encouraged to update immediately.”); Nat’l Cyber Sec. Ctr., supra; ITPro, supra.
[23] United States v. Carroll Towing Co., 159 F.2d 169, 173 (2d Cir. 1947).
[24] Cloudflare, supra; PHP Security Advisory, supra.
[25] Tenable, supra (removing public exposure lengthens the remediation window); Cloud Security Alliance, supra (a mitigation that removes public exposure counts toward meeting a deadline).
[26] TransUnion LLC v. Ramirez, 594 U.S. 413 (2021).
[27] In re Equifax, 362 F. Supp. 3d at 1321, 1325; see id. n.70 (noting that the defendants “do not seem to contend that the Plaintiffs have failed to establish standing”).
[28] Fed. Trade Comm’n, FTC Warns Companies to Remediate Log4j Security Vulnerability (Jan. 4, 2022), https://www.ftc.gov/policy/advocacy-research/tech-at-ftc/2022/01/ftc-warns-companies-remediate-log4j-security-vulnerability.
[29] Press Release, Fed. Trade Comm’n, FTC Takes Action Against GoDaddy for Alleged Lax Data Security for Its Website Hosting Services (Jan. 15, 2025), https://www.ftc.gov/news-events/news/press-releases/2025/01/ftc-takes-action-against-godaddy-alleged-lax-data-security-its-website-hosting-services; Hunton Andrews Kurth, FTC Issues Proposed Order Against GoDaddy and Provides Guidance on Security Practices, https://www.hunton.com/privacy-and-cybersecurity-law-blog/ftc-issues-proposed-order-against-godaddy-and-provides-guidance-on-security-practices (final order approved 3-0 on May 21, 2025); FTC Settlement Requires Disconnection of Hardware from All No Longer Supported Software, Data Protection Report (Feb. 2025), https://www.dataprotectionreport.com/2025/02/ftc-settlement-requires-disconnection-of-hardware-from-all-no-longer-supported-software (quoting the order).
[30] 23 N.Y.C.R.R. § 500.5(b), (c).
[31] N.Y. Dep’t of Fin. Servs., Press Release (Aug. 2, 2022), https://www.dfs.ny.gov/reports_and_publications/press_releases/pr202208021.
[32] Press Release, Coalition, Coalition Finds Organizations with Unresolved Critical Vulnerabilities Are 33% More Likely to Experience a Cyber Claim (May 16, 2023) (announcing the 2023 Cyber Claims Report), https://www.businesswire.com/news/home/20230516005410/en.
[33] Regulation (EU) 2024/2847 (Cyber Resilience Act), arts. 14(1), 14(2), 69(3), 71(2), OJ L, 2024/2847, 20.11.2024.
[34] Nat’l Cyber Sec. Ctr., supra.
[35] Cyber Resilience Act, supra, art. 3(42); Edward Machin, The Vulnerability Was Real; The Exploit Was Impossible; The EU Cyber Resilience Act Reporting Obligation Was … Unclear?, Ropes & Gray (Sept. 14, 2026), https://www.ropesgray.com/en/insights/viewpoints/102o10i/the-vulnerability-was-real-the-exploit-was-impossible-the-eu-cyber-resilience-a.
[36] Halborn, Explained: The BadgerDAO Hack (December 2021), https://halborn.com/blog/post/explained-the-badgerdao-hack-december-2021; DeFi Protocol BadgerDAO Exploited for $120 Million in Front-End Attack, The Block (Dec. 2, 2021), https://www.theblock.co/post/126072.
[37] Sansec, PHP2Shell: Critical PHP RCE (CVE-2026-109618, CVE-2026-109619) (Oct. 21, 2026), https://sansec.io/research/php2shell.
[38] 23 N.Y.C.R.R. § 500.17(a).
[39] PHP Security Advisory, supra; W3Techs, supra (27.0 percent of PHP sites on PHP 7 and 7.7 percent on PHP 5).
[40] McMillan, supra.
What kind of lawyer would I be without a disclaimer?
Everything I post here constitutes my own thoughts, should only be used for informational purposes, and does not constitute legal advice or establish a client-attorney relationship (though I am happy to discuss if there is something I can help you with). I can be reached via email at dlopezkurtz@crokefairchild.com on telegram @davidlopezkurtz on twitter @lopezkurtz and on LinkedIn here.
© 2026 David Lopez-Kurtz. All rights reserved. Permission is granted to quote brief excerpts of this post and to use it for academic, educational, and scholarly purposes, provided the use is properly attributed by crediting David Lopez-Kurtz as author, identifying this publication and the post title, and, where the medium allows, linking to the original. All other rights are reserved, including reproduction of the post in whole or in substantial part, commercial use or republication, and the preparation of derivative works. For any use beyond what is permitted here, or if you are unsure whether your intended use qualifies, please ask; I am generally glad to say yes. No part of this post may be used to train, fine-tune, or ground artificial intelligence or machine learning systems, or for text or data mining, without my express written permission.


