Tuesday, November 17, 2026
News

How do you check if your Optimizely snippet was infected?

The code came from Optimizely's own CDN, so allowlisting that host protected nobody. Report URI found 68 affected organizations with no Content Security Policy and 50 more whose policy let the whole attack through.

A security and privacy dashboard with its status

Open Settings > JavaScript in every Optimizely Web project and look for 1688360.cc, atob( or any code you did not write. Report URI found 188 infected snippets belonging to 178 organizations on November 16, and says the real number is likely higher. It reported every file clean by 23:11 UTC.

Which Optimizely product and which field were hit?

Optimizely Web Experimentation, and inside it the Project JavaScript field. Optimizely's status page scopes the incident to "Web Experimentation (Snippet Delivery)". Report URI found that the attacker put code in each project's Project JavaScript field and changed nothing else in the snippet. The file still came from cdn.optimizely.com, under Optimizely's certificate, from a host every affected site already trusted.

That field reaches further than any experiment does. Optimizely's own documentation calls Project JavaScript the snippet's "first point of execution", and it runs on every page. You edit it at Settings > JavaScript. Customers who serve the snippet from their own domain were caught as well, since the file is built from the same project.

How do you tell whether your own snippet served the code?

Search the field, not the file, and search for three things rather than one. Report URI saw three loader variants: a plain script tag pointing at 1688360.cc, the same URL hidden in base64 behind atob(), and an inline copy of the payload that ran through eval where the remote host was blocked. Grep for the domain alone and the second variant will not show up.

The timestamps narrow the window. Report URI first saw an injection at 04:21 UTC on November 16, a timestamp Infosecurity Magazine also carries, and reported every file clean as of 23:11 UTC the same day. The card skimmer was a later addition, at 13:43 UTC, so its advice is to treat cards entered that day as potentially exposed. Cards typed into hosted payment iframes from Stripe, Adyen and similar were out of reach; cookies without the HttpOnly flag were not.

Why didn't Content Security Policy stop it?

Because the malicious code came from a host every affected site already trusted. Report URI crawled 256 of the affected organizations' own hosts and graded what it found: 68 had no CSP at all, 50 had one that allowed both the payload and the data theft, 17 blocked the data going out but not the address swapping inside the page, and 24 blocked the payload outright. That leaves 118 organizations where nothing in place stood in the way.

Two common settings were the difference. 'strict-dynamic' trusts any script a trusted script inserts, which is what the Optimizely snippet was doing. 'unsafe-eval' let the inline fallback run even where the remote fetch was blocked, which Report URI reports happened at HSBC. Its verdict on the obvious defense is blunt: "Allowing cdn.optimizely.com was never going to help."

Optimizely's own guidance on that point changed this week. Its live CSP documentation, stamped 17 November 2026, has no 'unsafe-eval' anywhere and an example policy whose script-src ends 'strict-dynamic';. The same page archived on 9 October told customers that some features are not available "unless you include 'unsafe-eval'". Its Chrome extension page, stamped 8 November, still tells customers to allow it.

What has Optimizely said, and what has it left out?

It has published one status update and no advisory. The entry, titled "Web Experimentation Security Incident" and posted at 01:02 UTC on 17 November, says an outside party exploited an Experimentation vulnerability and used that to inject JavaScript into "some customer projects", and that the hole is fixed and the code removed. More than fifteen hours later it was still marked Monitoring rather than resolved.

What is missing matters more than what is there. The status page gives no count, no time window, no attacker domain and no instruction to customers, including none to inspect the Project JavaScript field the attacker wrote into. Optimizely named a mechanism, "the Experimentation account signup flow", only in a statement given to BleepingComputer, which also carries its remediation list and the line "We have notified all affected customers." Salesforce told the same outlet that as a precaution it is confirming Optimizely scripts are disabled across its preview and production instances, which is more than its supplier has published. The pattern repeats the Fireworks breach that exposed Notion AI chats.

What should IT teams do about the Optimizely snippet attack today?

  • Open Settings > JavaScript in every Optimizely Web project, plus every experiment's custom code, and search for 1688360.cc, atob( and anything nobody on the team wrote.
  • Rotate Optimizely credentials and API tokens, and review which users and integrations can write to a project. Optimizely says it reset access for affected user accounts, which is not the same as every account that can write.
  • Treat session tokens in cookies without HttpOnly as exposed for November 16 and rotate them, along with cards typed into your own forms rather than a payment iframe.
  • Read your own CSP for 'strict-dynamic' and 'unsafe-eval' first. Allowlisting your vendor's CDN is not the control you thought it was, and the same holds for every tag manager you load.
  • Ask Optimizely in writing whether your projects were affected, since its public notice gives customers no way to self-identify. Then monitor what runs in visitors' browsers, as you would check a Lovable app for leaking tables.

Sources