The search engine choice screen on Android—that prompt which elsewhere in Europe appears when setting up a new phone and asks users which engine they want as default—has silently vanished from Swiss devices. No announcement, no explanation: it’s just gone. Switzerland’s Competition Commission opened a preliminary investigation on Tuesday to understand what happened and, more importantly, to assess whether the sudden disappearance harms competition.

The Swiss affair is a warning bell that goes beyond a single national market. For years European regulators, following the Android case and the 2018 fine, have required Google to offer users a genuine choice among search engines. The mechanism is now part of the standard experience across most of the European Union. Its abrupt interruption in a country closely tied to the EU regulatory framework adds fuel to a wider debate: whoever controls the defaults on a device largely controls user behavior, and this applies not only to search but to any digital service, AI assistants included.

The stakes involve a company’s ability to channel billions of daily queries into its own engine, with direct consequences for the advertising market and data collection. But the Swiss probe comes at a moment when the concept of a ‘choice screen’ is about to extend well beyond search. With the rise of LLMs integrated into operating systems and mobile devices, the question ‘which AI do you want to use?’ could become the next regulatory battleground. If Europe already mandates search engine choice, tomorrow it might require choosing the default AI assistant or language model.

For organizations already evaluating on-premise LLM deployments, this trajectory adds an important piece. If regulators start questioning AI assistant defaults, the pressure grows for interoperable platforms where the user—or the enterprise—can replace the vendor-supplied model with a self-controlled one, perhaps running on local infrastructure. Control over interaction data then becomes a sovereignty requirement, not just a performance or latency issue. Those developing enterprise AI strategies face a scenario where regulatory compliance and the ability to choose the inference provider—including self-hosted ones—will be increasingly intertwined.

The disappearance of the choice screen in Switzerland, then, is not a trivial bug or a technical oversight. It’s a concrete test of whether antitrust constraints hold when public attention fades. And the Swiss regulator’s response will say a great deal about how authorities will tackle future AI gatekeepers, where default models risk becoming even more pervasive than search engines.