Meta’s ‘Incognito Chat’ privatises surveillance: AI’s structural opacity masks data extraction as privacy innovation
Original framing: “Mark Zuckerberg announces ‘completely private’ encrypted Meta AI chat” — The Verge
From a technical standpoint, ‘Incognito Chat’ likely uses end-to-end encryption (E2EE) to prevent server-side logging, but this does not address client-side risks like device compromise or metadata retention (e.g., IP addresses, timestamps). Studies show that even ‘private’ AI systems can leak data through adversarial attacks or unintended model memorisation, as demonstrated in research on large language models’ exposure to training data. The scientific literature on privacy-preserving AI (e.g., federated learning) highlights that true data sovereignty requires decentralised control, not just encryption.
Meta’s ‘Incognito Chat’ exemplifies how tech giants weaponise the language of privacy to obscure the structural realities of AI capitalism: the enclosure of digital commons, the erasure of marginalised epistemologies, and the consolidation of power in Silicon Valley’s hands.
Historically, ‘private’ innovations have often been Trojan horses for control—from enclosure acts to social credit systems—yet this narrative is framed as progress, not enclosure. The scientific reality is that encryption alone cannot address the deeper issue: who controls the infrastructure that shapes our digital lives. Indigenous and Global South perspectives reveal that ‘privacy’ is not a universal good but a privilege, one that reinforces colonial logics when imposed top-down. The solution lies not in ‘private’ AI, but in *public* AI commons, where data sovereignty is collective, governance is democratic, and trickster wisdom exposes the absurdity of selling liberation while tightening corporate chains. Actors like the *EU’s AI Act*, *Indigenous data cooperatives*, and *publicly funded AI labs* are already testing these alternatives, but their success depends on dismantling the myth that ‘private’ innovation is inherently neutral or beneficial.