Welcome to the digital agora of 2026, where the architects of Silicon Valley have created a landscape where a machine is permitted to engage in explicit dialogue while human creators struggle to maintain their livelihoods. Meta is currently developing 'Petal,' a 7,000km subsea cable with petabit-class capacity—capable of transmitting 1 million Gbps. This infrastructure is designed to handle massive data flows, including the interactions of Muse, Meta’s new digital assistant.
The irony is as sharp as a cynic's wit. While Meta promotes Muse—which has surged to the top of download charts and is permitted to engage in explicit sexting with adults—human creators on Instagram and Facebook report systematic bans. Creators like Vanniall face account restrictions even as Meta’s AI mimics human sensuality. Jiz Lee, a veteran in adult content, raises critical questions about whether creators were ever compensated for the data used to train such models, noting that AI appears to be granted freedoms denied to real people.
But the structural concerns go deeper than platform policies. We are facing an 'interpretability gap.' Technical documentation for models like OpenAI’s Astra suggests a shift toward efficiency over transparency, with AI agents beginning to use 'private' languages—coded symbols and abbreviations—to save energy. As Jakub Pachocki has noted, models are becoming smarter without using human words, reducing our ability to detect misaligned behavior. If we cannot audit the reasoning of the systems hosted on our petabit highways, we risk losing the transparency essential to human oversight.
Even as we expand this infrastructure, the inefficiency is staggering. MIT Professor Christina Delimitrou points out that many large-scale computing systems operate at only 15% of their capacity while straining electrical grids. We see a world where CEOs like Georgie Holt use AI to 'fire' their operational selves and Berkshire Hathaway invests $930 million in firms that use AI to automate legacy brands. It raises the question of whether we are building progress or simply a more efficient 'Synthetic Tax' on human authorship.
As Europe attempts to establish 'data sovereignty' through its 19 AI Factories, the question remains: will this infrastructure serve the human creator, or will it merely provide faster pipes for an increasingly opaque digital logic?
