Optimizely Script Attack Highlights Risks in Connected Customer Experiences

src-1996d8c6

A security incident involving Optimizely’s Web Experimentation platform showed how a compromise in a marketing technology supplier can reach customers across many organizations. On November 16, malicious code delivered through the platform was reported on websites associated with 178 organizations, including Salesforce, Zoom, GitLab, the BBC, CNN and HSBC.

Security researcher Scott Helme said attackers had inserted code into the Project JavaScript field within customer projects. The code was designed to steal payment card details, replace cryptocurrency wallet addresses, collect browser data and deliver iOS exploit kits. The incident put a particular spotlight on the scripts that personalization and experimentation tools can place on customer-facing websites: an organization’s own site may rely on code supplied or managed by an external provider.

The episode was one of several recent disruptions and security events affecting major technology providers and their customers. Microsoft 365 and Amazon Web Services experienced outages, while other incidents affected aviation communications, bank transactions and third-party suppliers. Together, the events underscored a practical issue for customer experience leaders: a service can fail or be compromised outside a company’s direct control, yet customers may still experience the consequences through its channels.

Marketing technology is part of the supply chain

Web experimentation platforms support the testing and delivery of changes to digital experiences. Their integration into a company website can also create a route through which a supplier-side issue reaches visitors. In the Optimizely incident, the reported malicious code was associated with customer projects, illustrating why organizations need to understand what third-party scripts can do and how changes to those scripts are controlled.

The reported range of intended actions—from payment-card theft to browser-data collection—makes the risk more than a matter of a disrupted website or an unsuccessful experiment. A compromised script could put visitors’ information or activity at risk. Helme described the event as a near miss and called for stronger protections. The incident details provided do not establish that every organization named experienced the same impact, or that each of the code’s intended functions succeeded.

This is a supplier-risk challenge as well as a security issue for marketing and technology teams. Organizations need visibility into the external code running on customer-facing pages, clarity about who can modify it, and procedures to respond if a supplier or integration is suspected of compromise. Those controls matter for experimentation and personalization tools because they operate close to the customer journey, where a technical change can affect what visitors see or how their browsers interact with a site.

Outages can interrupt service beyond the affected platform

The other incidents described this week show how dependencies can extend from a primary platform into communications and customer operations. Microsoft attributed its November 12–13 Microsoft 365 outage to several factors, including a software defect introduced in July, increased traffic after a Teams infrastructure change and reduced server capacity. Affected services included Outlook, Exchange Online and Dynamics Contact Center 365, along with elements of Teams and Copilot.

For businesses, the impact of an email or collaboration outage can reach beyond employee convenience. Service updates, customer enquiries, escalations and coordination with suppliers may depend on those channels. Separately, AWS reported authentication and connectivity problems in its US-East-1 region on November 15. Amazon Connect was among the affected services. AWS said the problems began at 3:05 a.m. PT; authentication issues were mitigated by 5:20 a.m., while some Network Load Balancer connectivity problems continued until 6:21 a.m.

These examples illustrate why backup arrangements need to account for dependencies beneath the customer-facing service. A contact center may be an alternative to email, for example, but access to it can still depend on identity or connectivity services. Organizations need to map those dependencies and test whether their contingency channels remain usable during a disruption.

Supplier incidents can expose customer information

On November 19, a ransomware group called n0n leaked data associated with customers of Visa, Mastercard and JPMorgan Chase after attacks on third-party service providers, according to the reported disclosures. Visa said a provider in the Asia-Pacific region was involved, while stating its own systems were not affected and that it had no evidence payment information, including card numbers, was exposed. Mastercard said a limited amount of information relating to some cardholders in a specific benefits program had been exposed. JPMorgan said the affected vendors held marketing data that may include customer names and email addresses, and said the incident did not disrupt its customer services.

Visa’s Fall 2026 Biannual Threats Report had warned that third-party agents were increasingly involved in incidents. For the January-to-June 2026 period, the report recorded a 62 percent increase in incidents involving third-party agents compared with the previous six months; those providers were linked to approximately 56 percent of compromised payment credentials identified in that period.

For marketing and CX leaders, the practical response is to treat supplier resilience, data access and customer communications as connected responsibilities. Teams should know what information vendors hold, how their integrations are monitored, and where customers can turn for legitimate help if normal channels fail. Clear, pre-established guidance can also help customers distinguish official support from fraudulent messages exploiting an outage. Resilience is not simply about preventing every incident; it is about maintaining secure, understandable service when a provider or dependency is disrupted.