Attackers turned a trusted website-testing tool into a malware delivery route on November 16, 2026. The Register reported that the Optimizely hack put malicious JavaScript on sites run by 178 organizations, after an unauthorized party exploited a flaw in Optimizely’s Experimentation product. Report URI researcher Scott Helme and his team found the code.
Report URI examined 1,611 Optimizely scripts and found that 188 of them carried malicious code. That works out to roughly one in nine. Our earlier report on the crypto address swap covered the wallet angle. This analysis looks at the mechanics, the numbers, and what the incident means for every team that loads someone else’s JavaScript.
The story matters beyond one vendor. The affected sites did not get hacked in the usual sense. Their own servers, databases, and login systems stayed untouched. The poisoned code arrived through a script they had chosen to load, from a vendor they had chosen to trust, in a format their security tools had no reason to flag.
Don't miss new tech stories on Google
Add Shattered once in the Google app and our stories show up in your news suggestions.
What Happened in the Optimizely Hack on November 16
According to the company’s own statement, Optimizely identified an incident in which an unauthorized party exploited a vulnerability in Experimentation and used it to add malicious JavaScript to some customer projects. The company said it blocked access to the vulnerable path, deployed a fix, revoked tokens, removed the injected code from affected projects, and notified affected customers.
Experimentation is the product that lets marketing and product teams run A/B tests and personalize pages without waiting for an engineering release. Customers drop a single snippet into their site, and the platform serves the rest. That convenience is the whole point of the product. It is also what made the attack so effective, because the snippet already runs on every page where it is installed.
Reports date the discovery to November 16, 2026, and coverage landed the following day. Help Net Security’s headline put the scale at 178 organizations and named the BBC, CNN, and Zoom among them. The sources reviewed for this article do not say how long the malicious code sat in customer projects before Helme’s team spotted it. Anyone claiming a precise exposure window is guessing.
The Numbers: 188 of 1,611 Scripts, 178 Organizations
Three figures anchor the incident, and they come from Report URI’s analysis as relayed by the outlets named below. The first is 1,611, the number of Optimizely scripts the team examined. The second is 188, the number that contained malicious code. The third is 178, the number of organizations tied to those scripts.
The gap between 188 scripts and 178 organizations is worth a second look. Some organizations apparently had more than one affected script, which fits how large companies run several Optimizely projects across brands, regions, or product lines. The sources do not break this down, so treat that reading as an inference, not a confirmed detail.
One caution on the 11.7% figure. Report URI’s sample of 1,611 scripts is not described in the sources as a random slice of all Optimizely customers. It may reflect the scripts the team could observe, not the whole customer base. Do not read the ratio as the of every Optimizely customer that was hit.
Which Organizations Were Reported Affected
Reporting on the incident names a roster that reads like a directory of household brands. Zoom, BBC, CNN, Salesforce, GitLab, IKEA, HSBC, Equifax, Air France-KLM, Ledger, and Peacock all appear in coverage. They span media, enterprise software, retail, banking, credit reporting, aviation, crypto hardware, and streaming.
That spread tells you something the attack surface. Experimentation tools are not a niche item. Any company that runs conversion tests, personalizes landing pages, or stages feature rollouts on its public site is a candidate customer. The same platform that tunes a news homepage also tunes a bank’s sign-up flow, which is why a single weakness reached so many different industries at once.
Being named in a report does not mean every visitor to those sites received malware. The sources establish that malicious code was associated with each organization’s Optimizely scripts. They do not publish per-site visitor counts, conversion rates for the attacker, or confirmed victim totals. Each affected company will have to determine its own exposure from its own logs.
Several of the listed names also sit in regulated sectors. Equifax and HSBC handle financial data, and Ledger sells hardware that people use to guard cryptocurrency. For those firms, the question shifts quickly from whether the code ran to which pages carried it and whether any of them touched regulated data. We return to that compliance angle below.
How Malicious JavaScript Rode Inside Optimizely Experimentation
The core of the attack is simple to describe. The attacker exploited a vulnerability in Experimentation, then used that access to add malicious JavaScript to customer projects. Optimizely projects can carry their own custom JavaScript, and the platform bundles that code into the script that the customer’s website loads. Put attacker code in the project, and the vendor’s own infrastructure ships it to every visitor.
Why the delivery path looked legitimate
Browsers and security tools judge a script largely by where it comes from. A file served from the vendor’s normal delivery network, referenced by a tag the customer wrote themselves, matches every allowlist the customer built. Nothing the request looked unusual. The code arrived through the front door.
This is the defining feature of a supply-chain attack . The target’s perimeter works exactly as designed. The weakness sits one step upstream, in a service the target treats as part of itself.
What Helme’s description tells defenders
Helme described the technique in plain terms, as Security Boulevard recounted . In his words, “The attacker wrote their own code into the Project JavaScript of customer projects, and the rest of each snippet was untouched.” The detail matters because it explains why a casual review would miss the change. A file that looks 99 percent normal passes a glance. Only a diff against a known-good copy, or a monitor that fingerprints the script, exposes the extra block.
What the Malware Could Do to Visitors
Reports on the payload describe a multi-purpose toolkit instead of a single trick. The malicious code could replace cryptocurrency wallet addresses, skim payment-card details, steal cookies, and capture screenshots. For some visitors it also delivered the Coruna and DarkSword iOS exploit kits, according to the reporting.
Each capability targets a different kind of loss. Wallet-address replacement aims at anyone to send crypto, since a pasted or displayed address can be silently swapped for one the attacker controls. Card skimming targets checkout pages. Cookie theft opens the door to session hijacking, which can bypass passwords and sometimes multi-factor prompts. Screenshots can reveal whatever a user happens to have on screen, from account balances to internal dashboards.
The iOS exploit kits raise the stakes. A skimmer steals what a visitor types. An exploit kit tries to compromise the device itself. The sources reviewed here do not say how many devices, if any, were successfully compromised, which kits worked against which iOS versions, or who built them. Until vendors or researchers publish that, the safest reading is that iPhone and iPad visitors were targeted, with outcomes unconfirmed.
The pairing of web skimming with device exploitation also hints at the attacker’s ambitions. Selling stolen card numbers is a familiar business. Spreading exploit kits through popular sites looks more like an effort to build access, and it suggests the operators saw a high-traffic injection point as worth more than a quick payout. We should be careful here, since motives are not established in any source. But the combination deserves attention from defenders.
Readers who held crypto or entered card details on any of the named sites around November 16 should check account activity and review recent transactions. Our guide to a cold storage wallet audit covers address-verification habits that blunt this kind of swap.
Does “World’s Biggest Websites” Hold Up?
The headline on The Register’s story calls the victims the world’s biggest websites. As a literal ranking claim, that is unconfirmed. No source in this reporting ranks the affected sites against global traffic lists, and 178 organizations is a count, not a leaderboard.
What the sources do establish is narrower and still serious. Popular, widely recognized websites were affected, and the 178 organizations include several that most internet users would know on sight. Headlines compress. Treat the phrase as shorthand for “very prominent sites,” not as proof that the largest sites on earth were all compromised.
This distinction matters for anyone writing internal incident briefs. Overstating the victim list invites easy rebuttals and erodes trust in the rest of the analysis. Stick to what the researchers measured: 188 malicious scripts out of 1,611 examined, tied to 178 organizations, with a named list of well-known brands among them.
Optimizely’s Response and What It Has Not Said
Optimizely’s public account lists five concrete actions. It blocked access to the vulnerable path, deployed a fix, revoked tokens, removed injected code from affected projects, and notified affected customers. Those steps match a standard containment playbook: close the hole, invalidate credentials the attacker might hold, scrub the damage, and warn the people exposed.
The statement leaves several questions open, and the table below separates what is confirmed from what is not.
Note the phrase “some customer projects” in the company’s own wording. Optimizely describes the compromise as limited to a subset of projects, while Report URI’s tally points to 178 organizations. Those two statements fit together, but only the vendor knows the true count of touched projects. Customers should not wait for that number before running their own checks.
What Optimizely and the Researcher Said
Optimizely’s statement, as The Register published it, opens with a plain admission: “Earlier today, Optimizely identified a security incident in which an unauthorized party exploited a vulnerability in Experimentation and used it to add malicious JavaScript to some customer projects.” The wording is blunt for a corporate notice. It names the product, the actor type, and the result, without hedging whether customers were affected.
SecurityWeek quoted the closing assurance from the same statement: “We’ve fixed the vulnerability and removed the injected code from affected projects.” That sentence covers remediation on the vendor side. It does not tell a customer whether visitors ran the code before removal, which is the question each security team now has to answer for itself.
Helme’s quote, shown earlier, supplies the technical counterpoint. The company says the problem is fixed. The researcher says the malicious code sat inside otherwise untouched snippets. Both can be true, and together they explain why detection took outside eyes. Optimizely’s own tooling served the files exactly as stored, so nothing on the delivery side looked broken.
Why Standard Browser Defenses Struggle Here
Teams often assume a content security policy or integrity check would catch this. In practice, most common controls assume the trusted vendor stays trustworthy. When the vendor’s origin delivers the bad code, a domain allowlist waves it through.
The table reflects our own assessment of how these controls behave, not a finding from the incident reports. The pattern is clear, though. Defenses that check where a script comes from miss attacks that change what the script says. Defenses that watch content, behavior, and outbound connections have a better chance.
This is also why Report URI’s role stands out. Its team found the malicious scripts by examining Optimizely scripts directly, which is exactly the kind of outside visibility most site owners lack. Teams that want broader coverage on the endpoint side can compare options in our review of endpoint protection platforms , though browser-side script monitoring is a separate layer that endpoint agents do not replace.
Historical Context: From Card Skimmers to Vendor Takeovers
Third-party script abuse is not new. Over the past several years, criminal groups running web-skimming operations, often grouped under the Magecart label, have repeatedly planted card-stealing JavaScript on checkout pages. Early waves hit individual shops. Later waves shifted upstream, compromising a widely used script or service so that one break-in reached thousands of sites at a stroke.
The 2024 Polyfill.io episode showed the same logic in a different form, where a popular script domain changed hands and began serving altered code to the sites that referenced it. SolarWinds in 2020 taught the broader industry that a trusted update channel can carry hostile code. The Optimizely case belongs to the same family, with one twist. The attacker did not buy a domain or poison a software build. They reportedly used a vulnerability in a customer-facing product to write into customer projects.
Other recent incidents on this site show the broader vendor-risk pattern. The Maileva breach exposed data tied to a shared service, as covered in our report on the La Poste Maileva hack , and a pair of intrusions at an AI inference provider reached customer chat records, as our coverage of the Fireworks AI breaches explained. Different vendors, same lesson: your risk includes everyone you plug into.
Market and Compliance Impact
No source in this reporting quantifies financial losses, stock moves, or customer churn tied to the incident, and we will not invent any. What we can describe is where the pressure will land.
First, vendors in the experimentation and tag-management space will face harder questions in procurement reviews. Security questionnaires already ask access controls and change logging. After November 16, buyers will likely press on who can edit customer-authored code, how edits get logged, and whether customers can lock a project against changes.
Second, compliance teams at affected firms will need to decide whether the incident triggers notification duties. For companies handling payment cards, the PCI DSS 4.0 rules on payment-page scripts, requirements 6.4.3 and 11.6.1, already require organizations to inventory and monitor the scripts that run on payment pages. An incident like this tests whether those controls worked in practice. Regulators in other regions may ask similar questions where personal data passed through affected pages.
Third, the reputational cost falls unevenly. Optimizely’s customers did not choose to be exposed, yet their brands carry the headline. The BBC, CNN, HSBC, and the others now appear in articles malware delivery, even though the reports place the vulnerability on the vendor side. That asymmetry is the central economic problem of third-party code: the customer bears the visible damage for a failure it could not directly prevent.
How Other Third-Party Script Models Compare
Optimizely is one of several ways a website can run outside code. It helps to compare the models by who controls the final script, since that determines where an attacker needs to break in.
Self-hosted scripts: Your team controls the file, your build pipeline signs it, and an attacker must breach you directly. Updates take effort, but integrity checks work well.
Vendor-hosted fixed libraries: The vendor serves a versioned file. SRI can pin it, and changes should be visible. The risk shifts to the vendor’s hosting and release process.
Vendor-hosted, customer-authored bundles: This is the Optimizely model. Customers write code in a dashboard, the vendor compiles and serves it, and the file changes whenever any authorized user edits a project. Pinning does not work, and every dashboard account becomes a path to production.
Tag managers and dynamic loaders: One container loads many scripts, so a single compromise can multiply. Governance depends on who can publish changes.
We make no claim here that any rival product was or was not vulnerable. The sources only Optimizely. The comparison shows that the vendor-hosted, customer-authored model concentrates risk in one place: the vendor’s authoring and publishing system. Any platform built that way deserves the same scrutiny.
What Security Teams Should Do This Week
If your site loads Optimizely, start with a direct check. Pull the current script, compare it against any saved copy, and review project-level custom JavaScript for code your team did not write. Optimizely for confirmation of whether your projects were among those affected, since the company says it notified affected customers.
Export the current Optimizely project JavaScript and review every custom block for unfamiliar code.
Check outbound requests from your pages to hosts you do not recognize, using browser developer tools or a script-monitoring service.
Rotate API tokens and review dashboard accounts, removing stale users and enforcing multi-factor authentication.
Ask the vendor for the date range during which your projects held injected code, then your own logs across that window.
Add script-change monitoring to checkout, login, and account pages first.
A simple fingerprint job gives you a baseline. This example hashes a vendor-hosted script so a scheduled task can alert on any change. Replace the placeholder URL with your own script path.
Because experimentation scripts legitimately change when marketers publish tests, expect noisy alerts at first. Route them to the team that owns the tests, and tag expected changes so real surprises stand out. Tighten your content security policy to restrict where scripts can send data, which at least limits exfiltration even when the script itself is trusted.
Credential hygiene matters too. Optimizely revoked tokens as part of its response, which suggests tokens were a concern. Scan your repositories and build logs for leaked keys using the approach in our secret scanning tutorial , and treat any vendor API key that ever sat in code as burned.
What Ordinary Visitors Can Do
Most readers cannot audit a site’s scripts, so the advice for individuals is practical and short. If you used a payment card on one of the named sites around November 16, watch your statement closely and consider asking your bank a replacement card. If you sent cryptocurrency that day, compare the destination address on your wallet’s confirmation screen against the address you intended before approving anything.
iPhone and iPad owners should install pending iOS updates, because reports describe exploit kits aimed at those devices. Sign out of sessions on affected sites and sign back in to invalidate stolen cookies. Beyond that, fresh bait is already circulating. Attackers routinely ride big news cycles, as shown by the fake GTA VI downloads carrying an infostealer , so treat any unsolicited “security update” email this incident with suspicion. Our security coverage hub tracks these campaigns as they develop.
Malwarebytes covered the incident as a supply-chain attack, and that framing is the right one for readers deciding how worried to be. The danger sits with the site’s supplier, not with anything the visitor did wrong. Good habits help, but they cannot fully offset a compromise this far upstream.
Five Predictions for the Weeks Ahead
The following are our own expectations, not reported facts. Treat them as informed judgment that later evidence may overturn.
Optimizely will publish more technical detail. A vendor that discloses an incident this prominent usually follows with a fuller account of root cause and timeline, and customers will demand it.
More named victims will surface. Companies that did not appear in the first reports will finish internal reviews and may disclose findings, widening the public list beyond 178.
Script-integrity monitoring will move from optional to expected. Procurement and audit teams will start asking for evidence of change detection on third-party scripts, especially on payment and login pages.
Vendors will add change controls. Expect experimentation and tag platforms to promote features such as approval workflows, publish logs, and per-user permissions for custom JavaScript.
Researchers will probe similar platforms. Helme’s team showed that watching served scripts at scale surfaces compromises quickly, and other researchers will point the same approach at comparable vendors.
What We Still Do Not Know
Several open questions limit how far any analysis can go today. The sources do not identify the attacker. They do not state how the vulnerability worked beyond saying it sat in Experimentation. They do not publish visitor exposure, financial losses, or the number of iOS devices that received exploit kits.
BleepingComputer’s coverage and other outlets will likely update as customers and the vendor more. Until then, the confirmed picture is this: an unauthorized party exploited an Experimentation vulnerability, malicious JavaScript reached projects tied to 178 organizations, 188 of 1,611 examined scripts carried bad code, and Optimizely says it has fixed the flaw and cleaned affected projects.
Frequently Asked Questions
What is the Optimizely hack?
An unauthorized party exploited a vulnerability in Optimizely’s Experimentation product and used it to add malicious JavaScript to some customer projects. Reports date the incident to November 16, 2026, and tie malicious scripts to 178 organizations.
How many sites and scripts were affected?
Report URI examined 1,611 Optimizely scripts and found 188 with malicious code. Those scripts were associated with 178 organizations. The sources do not provide a total count of individual websites or visitors.
Which companies were named?
Reports name Zoom, BBC, CNN, Salesforce, GitLab, IKEA, HSBC, Equifax, Air France-KLM, Ledger, and Peacock among the affected organizations. Being named means malicious code was linked to the organization’s Optimizely scripts, not that every visitor was infected.
What could the malware do?
Reports describe cryptocurrency-wallet address replacement, payment-card skimming, cookie theft, and screenshot capture. It also reportedly delivered the Coruna and DarkSword iOS exploit kits to some visitors. Confirmed outcomes per device are not published.
Has Optimizely fixed the problem?
Optimizely says it fixed the vulnerability and removed the injected code from affected projects. It also says it blocked access to the vulnerable path, revoked tokens, and notified affected customers.
Were these really the world’s biggest websites?
That phrase comes from a headline and is unconfirmed as a precise ranking. The sources establish that popular, well-known sites and 178 organizations were affected, nothing more specific.
What should website owners using Optimizely do now?
Review project JavaScript for code you did not write, rotate tokens, check outbound requests from your pages, and ask the vendor whether your projects were affected and for what dates. Add script-change monitoring on checkout and login pages.
Should visitors worry their cards or crypto?
If you paid by card or sent crypto on an affected site around November 16, review your statements and verify wallet addresses before confirming transfers. Keep iOS devices updated and treat unexpected security emails with caution.
The full story
This article is one source in a clustered incident — the cluster page carries the summary, timeline and every other outlet covering it.
