A vulnerability in website experimentation platform Optimizely let attackers inject malicious JavaScript into at least 178 widely visited websites, including major media, finance, and technology brands. The code stole payment card data, cookies, and screenshots, swapped cryptocurrency addresses for the attacker's own, and delivered exploit kits targeting older Apple iPhones and iPads. Optimizely says it has patched the flaw, removed the injected code, and reset affected credentials. Researchers found only 15 percent of affected sites had security policies strict enough to fully block the malicious script.

What Happened

On November 16, 2026, attackers exploited a vulnerability in Optimizely's Web Experimentation platform, which companies use to test and customize website content, to inject malicious JavaScript into the custom code field of customer accounts. Because the tampered script was delivered from Optimizely's own legitimate content delivery network under its own certificate, it reached visitors of at least 178 organizations' websites, including the BBC, CNN, Equifax, Ledger, and Zoom. Researchers at Report URI identified the campaign and found the payload was rebuilt 26 times over the course of the day. The malicious code was a multi-purpose kit: it stole payment card data, cookies, and screenshots, swapped cryptocurrency wallet addresses for the attacker's own, and on Apple iPhones and iPads, delivered zero-click exploit kits known as DarkSword and Coruna that can seize full control of outdated devices. Optimizely confirmed the incident, stating it identified and contained the issue, blocked the vulnerable path, removed injected code, ended active sessions, and revoked tokens. Researchers noted the attacker had tested similar code against Turkish websites back in September before this broader campaign.

Why It Matters

This incident demonstrates how deeply modern websites depend on third-party JavaScript they do not control. A typical homepage loads more than 20 separate JavaScript files from outside vendors, any one of which can become an attack vector if compromised. For leaders, this means vendor risk assessments must extend beyond contracts and compliance checklists to the technical question of what code a vendor's script is allowed to execute on your site. For defenders, the incident shows that trusted code is not permanently trustworthy, and runtime monitoring of third-party scripts matters as much as initial vetting. Analysts estimate roughly $7,500 in cryptocurrency was stolen through this campaign, alongside an unknown number of payment card records, underscoring that even a contained incident can carry real financial cost to end users.

What to Watch

Optimizely has published a report on its initial findings, and a fuller investigation with third-party experts is ongoing, so additional details about the vulnerability's root cause and total scope may still emerge. Organizations that use Optimizely should watch for further disclosures about the specific nature of the flaw, which could inform patching or configuration guidance for other experimentation platforms. It remains unclear exactly how many iOS devices may have been compromised through the embedded exploit kits, since researchers say they have no reliable way to measure that count.

What Defenders Should Do

Security teams should confirm with Optimizely or any similar experimentation vendor that injected code has been fully removed from their accounts and that affected tokens and sessions have been reset. Organizations should implement or tighten Content Security Policies that explicitly whitelist the domains allowed to serve JavaScript, since researchers found sites with strict CSPs were largely unaffected. Teams should also review their complete inventory of third-party scripts running on customer-facing pages and assess which of those scripts have the ability to read or modify sensitive user data such as payment forms.

The Larger Lesson

The incident reinforces a durable lesson in modern web security: a website's trustworthiness is only as strong as its weakest third-party dependency, and perimeter defenses built around a company's own code offer no protection against a compromised vendor script running with the same privileges as in-house developers. As AI tools make it easier for attackers to quickly write and redeploy malicious payloads, the gap between organizations that enforce strict content policies and those that don't is likely to widen.

Questions, corrections, or feedback about this story?Contact Daily Cyber News →

← Back to this Daily edition