Key Takeaways

  • Attackers injected malicious JavaScript into Optimizely Web Experimentation customer projects on November 16, 2026.
  • Report URI researchers identified malicious code in 188 of 1,611 Optimizely snippets linked to 178 organizations.
  • The incident highlights weaknesses in third-party script governance, including overreliance on trusted domains and certificates.

A third-party JavaScript compromise involving Optimizely Web Experimentation caused prominent websites to deliver malicious code to visitors on November 16, 2026. CNN, BBC, Zoom, Salesforce, and GitLab were among the named sites affected, although there is no evidence that attackers separately breached each company's infrastructure.

Attackers targeted a shared dependency rather than breaking into every customer environment. Once malicious code entered Optimizely customer projects, browsers treated the resulting scripts as legitimate because they came through Optimizely's infrastructure and certificate.

According to Malwarebytes, Report URI researchers examined 1,611 Optimizely snippets and found malicious code in 188 of them. Those snippets were associated with 178 organizations. The findings illustrate the potential reach of a compromise involving a widely deployed web service: one successful intrusion can create exposure across many otherwise unrelated brands.

The payload was not limited to one form of theft. BleepingComputer reported that it could replace cryptocurrency wallet addresses, collect cookies and screenshots, and skim card numbers and security codes entered into website forms. It also attempted to deliver the Coruna and DarkSword iOS exploit kits.

In practical terms, visitors could face different risks depending on what they did while an affected script was active. Someone who merely opened a page may have had different exposure from someone who entered payment information, accessed an authenticated session, or copied a cryptocurrency address. Organizations investigating the event will need to determine which pages loaded affected snippets, when those snippets were active, and what information users could enter on those pages.

A valid TLS certificate establishes an encrypted connection to a domain, but it does not prove that every script supplied by that domain is safe. If attackers gain control of a trusted third-party component, browsers can securely download malicious code directly through authenticated channels.

Detection was also a moving target. CISO Voice reported that the campaign's payload was rebuilt 26 times during the day. Frequent rebuilding can change hashes and other indicators, reducing the value of controls that depend only on matching a previously observed file. When the trusted supplier is also the delivery mechanism, conventional domain-based blocking offers little help.

Optimizely said it blocked the vulnerable signup route, removed injected code, revoked tokens, ended active sessions, and reset affected access. Those actions address immediate persistence and account-access concerns, but customers still have their own investigative work. Necessary steps include reviewing script inventories, identifying affected project IDs, preserving browser and content-delivery logs, and checking whether sensitive forms were present on impacted pages.

Subresource Integrity (SRI) can help browsers verify that an external script matches an approved cryptographic hash. Its use can be difficult with experimentation tools whose code changes regularly, however. Content Security Policy (CSP) can restrict permitted script sources, but a basic allowlist may not stop code delivered from an already approved third-party domain. More restrictive policies, script pinning, self-hosting where practical, and tightly controlled change processes can reduce exposure.

Marketing, analytics, experimentation, chat, payment, and personalization scripts execute inside the browser's trust boundary. Security reviews therefore require treating them as production code rather than incidental website add-ons.

Removing every external script is rarely realistic for modern web operations. A workable approach is to document ownership, limit privileges, monitor changes, establish rapid disablement procedures, and rehearse how a compromised dependency would be removed across multiple sites. The Optimizely incident demonstrates how trust inherited from a major brand or valid certificate can rapidly become an attacker's distribution channel.