Introduction
On November 16, 2026, 178 of the world's most recognized organizations, Zoom, BBC, Salesforce, GitLab, IKEA, HSBC, Equifax, and Air France-KLM, discovered they were simultaneously serving malicious code on their websites. Not one had been directly breached. The attack had traveled through a tool they all used and trusted: Optimizely, the widely adopted SaaS experimentation platform used to run A/B tests and power online personalization. The malicious code was served directly from the vendor's own CDN, making detection all the harder.
A Legitimate Mechanism Weaponized Against Customers
Optimizely gives clients the ability to deploy JavaScript snippets directly onto their pages through a dedicated field called "Project JavaScript." That is precisely the vector the attacker exploited: by gaining unauthorized access to customer accounts through a vulnerability in the sign-up process, they injected malicious code into 188 snippets across 178 organizations. The code silently loaded a script from a command-and-control domain registered just one week earlier, then activated multiple payloads in parallel. Security researchers who documented the incident named the campaign OPTIMJACK.
Four Attack Vectors, One Injection
What makes this incident particularly alarming is the sheer breadth of objectives pursued simultaneously. The deployed code carried out four distinct operations at once: silently swapping crypto wallet addresses displayed on pages to redirect transactions to the attacker; harvesting payment card data entered into forms; stealing session cookies and capturing screenshots for reconnaissance; and redirecting iOS devices running older versions to the Coruna and DarkSword exploit kits, as documented in post-incident analyses.
This versatility reveals a meticulously planned attack, deploying multiple monetization techniques in parallel to maximize yield during the exposure window.
A Sixteen-Hour Window, a Growing Toll
The malicious scripts were active from 04:21 UTC, with cleanup beginning around 20:40, roughly sixteen hours of exposure. Over that period, the cryptocurrency address-swapping scheme generated a documented haul of approximately $7,000, a figure that looks negligible given the scale of the brands affected, but is explained both by the brevity of the window and by the mechanics of address substitution: it only pays out when a visitor actually executes a crypto transaction during their session. Researchers who covered the incident were explicit: that figure represents only the immediately quantifiable fraction. Payment card skimming, by contrast, produces delayed effects whose true scope will take several weeks to fully measure.
The real magnitude of this incident is not monetary: it is that 178 organizations were simultaneously exposed without a single flaw in their own systems being exploited.
The Core Problem: Implicit Trust in SaaS CDNs
This incident illustrates a reality that security teams understand in theory but struggle to act on in practice: every third-party SaaS tool integrated into a website carries an implicit layer of trust. Scripts served from a recognized vendor's official domain naturally bypass user vigilance, and sometimes the vigilance of detection tooling as well.
The post-incident analysis makes for sobering reading: of 256 domains examined, only 24, just 15%, had a Content Security Policy strict enough to have blocked the payload. The remaining 85% had no technical barrier capable of preventing the injected code from executing.
What This Means in Practice
For IT and security teams, the Optimizely incident confirms three priorities that are too often deferred. First, Content Security Policy is no longer optional: it is one of the few defenses capable of neutralizing this type of client-side injection. Second, SaaS integrations must be included in third-party risk management programs, on equal footing with network access points and exposed APIs. Third, monitoring the integrity of third-party scripts must become a standard component of application observability, not remain a blind spot.
Trusting a SaaS vendor is not a security guarantee. It is a risk that needs to be actively managed.

