On November 18, Optimizely said an unauthorized party exploited a flaw in the signup flow for Web Experimentation to impersonate users on targeted customer accounts and insert JavaScript into their projects. Two days earlier, Report URI had counted 188 infected snippets associated with 178 organizations among 1,611 Optimizely snippets it examined. The count refers to organizations whose pages used infected snippets, not confirmed individual victims. Report URI’s technical account and Optimizely’s incident update describe the findings and the company’s explanation.

What Optimizely said caused the incident

Optimizely attributed the access to a flaw in the Web Experimentation signup flow. The company said an unauthorized party used it to impersonate users on targeted accounts and add malicious JavaScript to customer projects. Optimizely said the incident was limited to Web Experimentation, the flaw had been fixed, the injected code removed, and additional protections added.

The route matters because the code was added to projects that supplied scripts for customer pages. Report URI identified the injection point as Project JavaScript (projectJS). Its review described three loader forms: a plain script tag, a URL concealed in base64, and an inline payload using eval if a remote script was blocked. These were observed variants; the report did not say every infected snippet used all three.

How one Optimizely script reached many organizations' pages

Code Bear’s English-language explainer uses diagrams to trace how an Optimizely snippet loaded across customer pages

A snippet is a piece of code a site loads to use a service. When a shared service’s project script changes, pages using that script can receive the change through the same delivery path. Here, Report URI found malicious code in Optimizely project snippets associated with 178 organizations. The count describes the reach of those snippets; it does not count visitors or establish that every organization or visitor suffered data theft.

Treat every script that runs in a visitor’s browser as part of your site’s attack surface: control who can change it and restrict what it can run and send.

What the code was designed to do

Report URI said the payload could replace cryptocurrency wallet addresses in page content and other locations, and collect payment-card details entered into a site’s own form fields, including card numbers, expiration dates and security codes. Hosted payment iframes were outside the card-skimming path it described.

The code could also send certain cookies not marked HttpOnly and page screenshots when it detected wallet- or payment-related wording. Those are reported capabilities of the payload, not a tally of records stolen from visitors.

For Apple devices, Report URI described routing Coruna to iOS 13–17, DarkSword to iOS 18.x, and a separate chain on iOS 26 that fingerprinted devices. Infobae reported that MATT Research had not confirmed a successful infection of a real device, and that Reuters could not confirm the total number of victims. Infobae’s report covers that distinction between exploit delivery and confirmed outcomes.

What the reported scope and cleanup show

Report URI said the infected files it tracked were clean by 23:11 UTC on November 16. In its November 18 update, Optimizely said it had removed injected code and that no malicious code remained in its systems. These are separate accounts of the cleanup: one describes files Report URI observed; the other describes Optimizely’s systems.

The same incident analysis points to a practical distinction in Content Security Policy (CSP), a set of browser rules governing what a page’s code can load or send. Among 256 hosts assessed by Report URI, 24 had a policy that blocked the payload, while 17 allowed the payload but blocked data theft. A policy that restricts outbound data can limit exfiltration without necessarily stopping code from changing information on a page.

Report URI advised site operators to review Optimizely Project JavaScript and experiment code, check account write access and tokens, rotate credentials, and use restrictive CSP controls. For visitors, it recommended updating Apple devices, checking wallet addresses before transfers, and monitoring card statements.