This vulnerability reveals a fundamental flaw in SAPUI5's security model: the framework treats content adaptation privileges as a proxy for script execution trust. When key users modify application behavior through the adaptation layer, their inputs bypass the sanitization that would be applied to ordinary user input. Because these adaptations persist, a single malicious key user account creates a stored XSS vector that compromises every subsequent user who opens that application — automatically, without any social engineering required per victim.
The attack economics here are fundamentally different from reflected XSS. One compromised key user account generates persistent payloads that pay dividends across your entire user base. This transforms the key user role into a high-value target — far more valuable than a standard user account, because the blast radius extends to everyone consuming the adapted application.
The architectural problem is that content adaptation is marketed as a legitimate business capability, so organizations widely distribute key user privileges across the enterprise. SAPUI5's framework treats this privileged role as a trust boundary rather than a high-risk attack surface, meaning the persistence mechanism assumes 'key user output is safe' as a design principle. That assumption is precisely what creates the vulnerability.
You need to audit your content adaptation layers immediately. Identify which applications have key user customizations, map the victim pool (every user who accesses those applications), and treat any key user compromise as a potential mass-compromise event rather than a limited account takeover. The persistence mechanism means a malicious adaptation doesn't require ongoing access — one injection persists until the adaptation is removed or the application is rebuilt.
The supply chain dimension compounds this: key user privileges are often delegated to implementation vendors and system integrators. Your adapted applications may contain payloads authored by parties you cannot audit. Assume that any content adaptation layer is a potential injection surface and instrument detection accordingly.