Optimizely’s delivery network carried attacker-written JavaScript in 188 scripts linked to 178 organizations, security company Report URI says. The affected inventory included websites associated with Zoom, CNN, Salesforce and HSBC. Visitors could encounter malicious code through a service those websites had already chosen to trust, without needing to follow a suspicious link to an unfamiliar website.
The findings concern code delivered through websites. They do not establish that 178 organizations’ internal networks were breached. That distinction matters as The Hormuz Letter’s breaking post spreads a much shorter account of the attack. Report URI’s count came from 1,611 scripts it checked, and it says the full affected population is probably larger.
The trusted delivery was working
Optimizely Web Experimentation helps customers test changes to their websites. Its scripts can include customer-written Project JavaScript. Report URI found the attacker’s code in that field, while the surrounding script remained intact. Optimizely’s legitimate content-delivery network then served the modified material. The researchers explicitly distinguish this from poisoning the network’s cache with a fake file.
That explains why a familiar address and a valid security certificate offered little reassurance. HTTPS protects a connection to the server; it does not certify the intentions of whoever last edited the delivered script. A browser can receive a file securely from the correct supplier and still receive dangerous code.

Once a website permits an external script to run, that script operates inside the page. The company hosting the script does not become a separate protective container around it. Code can interact with the page’s accessible content and form fields. Renting a small website function therefore delegates access that is much broader than the function’s friendly description might suggest.
What the malicious code could reach
Report URI describes cryptocurrency-address substitution, a payment-card skimmer, conditional cookie and screenshot collection, and delivery of iOS exploit kits. These are documented functions and delivery observations, rather than a confirmed number of successful device compromises. The report does not supply a final card-exposure or loss total.
The payment distinction is crucial. The skimmer targeted cards entered into a website’s own fields. Cross-origin hosted payment frames, such as those supplied by Stripe or Adyen, were outside its direct reach. Browsers normally prevent a page from reading fields inside such a frame. That boundary protects a particular set of inputs; it does not stop malicious code from altering other accessible parts of the surrounding page.
The wallet snapshot also needs its limits. Report URI recorded about $7,040 received by attacker-linked wallets as of 19:25 UTC on November 16. That is a dated observation, not a tally of every victim’s losses. Cookie and screenshot theft in the analyzed code depended on an address swap near payment-related wording, rather than happening to every visitor.
The access question remains open
Scott Helme, who published the investigation with Report URI, identified the unanswered question: “We don't know how the attacker got write access to so many unrelated customer projects.” His report discusses possible explanations without settling on one. A description of the payload is not yet a full account of how the attacker acquired that reach.
Optimizely says an unauthorized party exploited a vulnerability in Experimentation. Its public response says the flaw was fixed, injected code removed and affected user accounts secured, with additional protections added. The company’s statement on X also says customers were contacted directly. Those are the supplier’s assertions about its response; they leave the detailed access mechanism undisclosed in the materials reviewed here.
Report URI’s November 17 update said all the infected files it was monitoring had been cleaned by late Monday. Optimizely’s incident record remained in monitoring Tuesday evening. An open incident record does not establish that malicious code is still being served. Equally, removing a script does not quantify what happened while it was available.
A browser rule stopped part of the attack
Report URI says restrictive script policies protected 24 sites. Other sites blocked data leaving the page while still allowing address substitution inside it. Content Security Policy lets a website restrict which resources a page loads and where it sends data. The protection depends on the actual rules: approving a supplier’s hostname is not an inspection of everything that supplier later delivers.
Report URI sells monitoring and browser-policy products, a commercial interest readers should know when weighing its recommendations. Its underlying findings nevertheless put a concrete question to website operators and their suppliers: who can change the code running in a visitor’s page, and what record proves that a change was authorized? That question also belongs in the wider discussion of automated attacks, without assuming this incident used AI.
What we expect next
We expect the next consequential disclosures to concern project-access records and actual visitor impact. Optimizely’s fix addresses the immediate delivery problem, while its published account still leaves the cross-customer write path unexplained. A technical account of that path would let customers assess whether their own controls could have contained it. Evidence from affected organizations will be needed before anyone can turn the script count into a reliable total of stolen cards, compromised devices or losses.
Common questions
What happened in the Optimizely attack?
Report URI found malicious JavaScript in 188 scripts associated with 178 organizations during the November 16, 2026 incident.
Were 178 organizations’ internal networks hacked?
That is not established. The reported count concerns compromised website scripts, rather than 178 proven internal network breaches.
How many Optimizely scripts did Report URI check?
Report URI checked 1,611 scripts and says its observed affected population may undercount the full incident.
Did Optimizely confirm a vulnerability?
Yes. In its public incident response, Optimizely says an unauthorized party exploited an Experimentation vulnerability and that it has fixed the flaw.
Has Optimizely explained the cross-customer access?
Its statements reviewed on November 17 do not explain the detailed mechanism for changing multiple customers’ projects.
Did hosted Stripe or Adyen payment frames expose their card fields?
The analyzed skimmer could not directly read those cross-origin hosted frames. It targeted fields in the website’s own page.
How much money was stolen?
No complete loss total is established. Report URI’s roughly $7,040 wallet snapshot was recorded at 19:25 UTC on November 16.
Were iPhones and iPads successfully compromised?
Report URI observed exploit-kit delivery, but it did not establish a total of successfully compromised devices.
Were cookies and screenshots taken from every visitor?
No such finding was reported. The analyzed collection logic depended on an address swap near payment-related wording.
Was Optimizely’s delivery cache poisoned?
Report URI says no. The delivery network served malicious code that had been added to customer projects.
Had the malicious scripts been cleaned by November 17?
Report URI’s November 17 update said its monitored infected files were clean by late November 16.
Why was the security incident still listed as open?
Optimizely’s incident remained in monitoring Tuesday evening. That status alone does not mean malicious code was still being served.
Does HTTPS guarantee a downloaded script is safe?
No. HTTPS protects the server connection, rather than certifying the intentions or safety of the delivered JavaScript.
Was the attack linked to a named country or AI tool?
No such attribution was established in the sources reviewed for this November 17 article.



