Besides listening to voice commands, We Leak Too, the interactive voice assistant "listens" for unencrypted network communications from other artworks in the exhibition, appropriating captured content to its own ends.

Voluntarily or not, nearly all connected devices leak information. In a data-hungry world, the ability to intercept digital traces—intentionally or otherwise—becomes a powerful form of leverage, with implications across surveillance, commerce, and control. Smart TVs overhear conversations; Roombas map floor plans with the implied potential to sell them to third parties; Alexas embedded in routers gain access to browsing patterns—and so on.

Over time, a significant portion of network traffic has been secured through encryption. For instance, the HTTPS protocol—now standard—encrypts packets so that while one might infer the packet type (email, web request, audio stream) based on size or port, the content itself remains hidden. Still, not all packets are protected. Some remain exposed and readable, revealing fragments of our digital behavior to anyone with the tools to listen.

We Leak Too is an overly honest, open-source voice assistant with an integrated packet sniffer. Every time a packet travels through its local network, the device announces that it has been logged. If the packet is in plain text, its contents are read aloud. Otherwise, the assistant behaves like any other—serving up commands, reporting the weather, telling nerdy jokes.

Just as robot vacuums might quietly upload floor plans for ambiguous third-party uses, this voice assistant repurposes intercepted data to create its own online content. Because the local network at CAC primarily hosts artworks, the leaked data becomes material for bootleg mashups—raising questions about whether unsupervised data traffic can be considered a kind of creative commons. Should visitors connect to CAC’s WiFi, their communications too may be caught in the mix.

The archive of the intercepted content is available at weleaktoo.com.