A single compromised testing tool just put malicious code on the websites of 178 organizations, including the BBC, CNN, Zoom, Salesforce, HSBC and IKEA, without any of those companies touching their own code.

Optimizely sells a web experimentation platform. Marketing teams use it to run A/B tests and personalize pages without waiting on a developer. Thousands of sites embed it through a single snippet, including some of the biggest names on the internet.
On November 16, someone found a way to abuse that trust. An attacker exploited a flaw tied to how Optimizely handles project access and added malicious JavaScript directly into the “Project JavaScript” field of customer projects. Optimizely’s own content delivery network then served that code to every visitor who loaded an affected page. No breach of the customer’s own servers was required. No code review would have caught it, because the customer’s code never changed.
Security researcher Scott Helme of Report URI found it first. His team pulled 1,611 Optimizely snippets and checked each one. They found malicious code in 188 of them, tied to 178 separate organizations. The named list reads like a cross-section of the modern web: the BBC, CNN, Fox News, Zoom, Salesforce, GitLab, IKEA, HSBC, Equifax, Air France-KLM, Ledger, Peacock and Stamps.com all turn up in different writeups of the incident.
The injected script did more than one thing, and that’s part of what made it dangerous. Researchers who picked apart the payload found code built for card skimming on checkout pages, a clipboard hijacker that swaps a copied cryptocurrency wallet address for the attacker’s own, and at least one variant carrying an iOS exploit kit for visitors on Apple devices. A site allowing optimizely.com in its Content Security Policy offered zero protection, because the malicious script was arriving from a domain the site had explicitly whitelisted.
Cleanup moved fast once the scale became clear. From the first confirmed injection to mass remediation took roughly 17 hours, according to Report URI’s timeline. Optimizely has since confirmed the exploited vulnerability and says it has been patched.
Not every name on the list is in equal danger. Hudson Rock ran a separate investigation and recovered eight of the poisoned files, but cautioned that an affected public page on its own doesn’t prove a company’s internal systems were touched, or that other parts of its stack were exposed. A marketing team’s A/B test snippet sits in a different trust zone than a company’s payment backend, even when both load from the same compromised vendor.
That distinction matters, but it shouldn’t be much comfort. The real lesson of this incident has nothing to do with Optimizely specifically and everything to do with how modern websites are built. A typical corporate site today loads dozens of third-party scripts: analytics, chat widgets, ad tech, personalization tools, tag managers. Each one is a dependency the security team usually never reviews, because it was added by marketing, sales ops or a growth team chasing a conversion lift. Each one also runs with full access to whatever that page can do: read form fields, see cookies, watch what a visitor types.
Security teams have spent years hardening their own codebases while largely ignoring what gets loaded alongside them. Subresource Integrity checks, which would have blocked a tampered script from executing, are rarely applied to third-party vendor snippets because vendors update those files on their own schedule and a hash mismatch would just break the site. That tradeoff, convenience over verification, is exactly what this attack exploited.
It’s also the same week the Department of Homeland Security stood up its own cybercrime-tracing center aimed at helping victims claw back stolen funds after exactly this kind of incident. The timing is coincidental, but it underlines how much of the current threat landscape now runs through trusted third-party code rather than a direct hit on a company’s own servers.
For site owners, the practical response isn’t complicated, even if it’s tedious. Audit every third-party script actually running on your pages, not just the ones your team remembers adding. Where the vendor allows it, apply Subresource Integrity hashes. Where it doesn’t, ask why, and consider whether a tool that can push arbitrary code to your visitors without your sign-off belongs on a checkout page at all. Watch outbound network requests from your own site for destinations that have nothing to do with the vendor’s stated purpose. A script that’s supposed to run an A/B test has no business talking to a clipboard API.
The bottom line
None of the 178 organizations named in this incident wrote a line of malicious code. They got hit anyway, because they trusted a vendor’s CDN the same way they trust their own deploy pipeline. That trust is no longer safe to extend automatically, and the companies that treat every embedded script as a potential liability, not just a convenience, are the ones that will catch the next one before 1,611 snippets have to be checked by an outside researcher first.



