The Internet lies now amidst a new cycle of disruption due to the mass adoption of artificial intelligence (AI). As a research field, AI even predates the Internet; however, the technology just reached the necessary maturity and availability of resources and capital. In fact, AI does not think. It just extracts correlations from large amounts of data and training material and then uses them to make predictions or generate new content. The result statistically imitates the production of a human brain but is inherently bound to contain mistakes. It also replicates societal biases, and it is arguable whether its output constitutes original creation. Indeed, AI will bring forth societal and economic changes. It could thus further erode the digital sovereignty of countries that rely on foreign AI providers and fragment the Internet into ecosystems shaped by competing AI models. Europe has tried to address these risks via regulation (AI Act); adopting the open-source paradigm could also help. In the end, AI could even fragment reality, enticing us into virtual relationships with simulated humans. We must build our ability to use AIs as tools rather than depend on them.