In the Optimizely hack on 16 November 2026, malware ran through the A/B testing snippet on the websites of 178 organisations, including the BBC, CNN and Salesforce, according to Report URI. It skimmed card data and cookies, delivered from the tool's own trusted domain.
Key facts
- According to Report URI, 188 Optimizely snippets from 178 organisations carried malware on 16 November 2026, out of 1,611 checked.
- The code swapped crypto addresses, skimmed card data from on-site forms, sent cookies and screenshots out and loaded exploits for older iPhones.
- Optimizely says the Web Experimentation flaw is fixed; per Report URI only 24 affected companies had a CSP that would have stopped the attack.
Researched and written with AI, to Luigi's brief, positions and editorial rules. Automatically checked against the sources. How I work →
You know the one. The testing tool went in three years ago, a single line in the header, and nobody has asked about it since. It runs on every page, checkout included. On Sunday, that one line served malware for 178 organisations.
What happened in the Optimizely hack
On 16 November 2026 attackers wrote code into projects in Optimizely Web Experimentation, the A/B testing tool many big brands rely on. According to BleepingComputer, they abused a flaw in the Experimentation account signup flow. The code then shipped the normal way, from cdn.optimizely.com: trusted domain, valid certificate.
- Report URI (security researcher Scott Helme) found 188 infected snippets from 178 organisations out of 1,611 checked. Named sites include the BBC, CNN, Zoom, KLM, Air France, Salesforce, GitLab and UNICEF USA.
- Report URI saw the first injection at 04:21 UTC on the BBC; most of the cleanup happened between roughly 20:40 and 21:10 UTC. The payload was rebuilt 26 times that day.
- It swapped crypto wallet addresses (about $7,040 taken, per Report URI), skimmed card data from on-site forms from 13:43 UTC, sent cookies and screenshots to the attackers and loaded exploit kits on older iPhones and iPads.
Optimizely says on its status page that the flaw is fixed, the code removed and affected accounts secured. Salesforce has disabled Optimizely scripts as a precaution. Back in February 2026, Optimizely had already confirmed a voice-phishing attack on internal systems, SecurityWeek reported.
Your A/B testing script is an admin login, not a plugin
This is the part that matters to me. A testing tool is allowed, by design, to change anything on your page: copy, buttons, forms. That is what you pay for. So whoever can write into the tool writes straight into your website. No deploy, no IT review, nobody on the team notices.
And it wasn't a one-off. Report URI says the same attacker infrastructure ran through scripts from Kissmetrics, StorifyMe and Minute Media between 13 and 16 November. Analytics, content widgets, a media player. Marketing tools, every one.
It's a pattern I've watched in marketing for years: we run best of breed, one tool per job, and every agency brings its own snippet. Now add AI chatbots, AI personalisation and agents that set up tests on their own. The way I see it, that means more third-party code on the page and more places with write access to it. AI doesn't fix that foundation. It widens the cracks.
The catch: protection existed, few had it switched on
A Content Security Policy (CSP) controls where a script on your page may send data. Report URI checked the setups of affected companies: in only 24 cases would the CSP have stopped the payload, and 68 had none at all. Payment fields inside hosted frames, such as Stripe or Adyen, were not affected, according to Report URI. For the record, i4u.business runs a strict CSP that blocks third-party scripts unless they are explicitly allowed.
What I'd do now
- Script inventory: list every third-party script on your site with an owner on the team. Anything nobody can explain, or with no live test, goes.
- Check the testing tool: Optimizely customers should review the Project JavaScript field and custom experiment code (Report URI names the domain 1688360.cc and "atob(" as traces), audit users and API tokens, and rotate credentials.
- Sort out CSP with IT: restrict connect-src first, so a foreign script can't ship data to any server it likes.
More on the same gap in the stack: the FBI warning on ARTEX and forgotten integrations, Google's report on attacker AI and vibe-coded apps leaking customer data. If you're building landing pages with AI or adding an AI chatbot to your website: ask first who owns the script.
Bottom lineEvery marketing script on your site is an employee with a master key, and most of them were never hired.
Got AI news? Submit news →
FAQ
What happened in the Optimizely hack?
On 16 November 2026 attackers wrote code into Optimizely Web Experimentation projects. Per Report URI it ran on websites of 178 organisations and skimmed card data and cookies, among other things.
Which websites were affected?
Named sites include the BBC, CNN, Zoom, KLM, Air France, Salesforce, GitLab and UNICEF USA. Report URI expects the real number to be higher.
What should Optimizely customers do now?
Check Project JavaScript and experiment code for foreign code, review users and API tokens, rotate credentials and deploy a CSP with a restricted connect-src.
Source
Report URI: World's largest brands infected with malware via Optimizely
More sources
- Optimizely Status – Stand 17.11.2026 01:02 UTC: Lücke behoben, Code entfernt, Konten gesichert, Status Monitoring
- BleepingComputer – 16.11.2026: Lücke im Anmeldeprozess für Experimentation-Konten, betroffene Websites, Statements Optimizely und Salesforce
- SecurityWeek – Februar 2026: Optimizely bestätigt Vishing-Angriff auf interne Systeme




