Chapter 6 of Bad Computers by Michael Veale and Robert Gorwa

Comms: Distributing Tutorials and AI Output

The Internet is all about communication, and every element of the stack covered within this book so far is involved in transmitting information to and through machines. But some parts of the Internet have been explicitly designed for humans to get in contact with other humans: messaging apps, social networks, video and file-sharing services. Let’s call them ‘comms platforms.’

Comms platforms are somewhat distinct in the AI stack in other ways. Perhaps more than any other layer, they recur in multiple distinct roles throughout the entire pipeline of model misuse. In this chapter, we begin by looking at how comms platforms help animate and coordinate malicious activity by hosting tutorials, communities of potential and actual abusers, as well as ads and marketplaces for services used by these abusers. We then examine how these platforms are used to intermediate between AI generation and the victims of various types of malice. This might be through more private means, such as direct messages sent by a scammer, or more public means, such as the circulation of non-consensual sexual forgeries. Although this initially may seem like just a content moderation problem — something which has been written about with some depth when it comes to AI-generated media — we explore the various constraints that can make this a much more difficult problem than often acknowledged in the extant policy conversation. Throughout, we focus on the new issues raised by multipurpose generative AI systems: how might these comms platforms respond?

Download chapter 6 as a PDF (pages 147–175, 1.8 MB).

How to cite

Michael Veale and Robert Gorwa, Bad Computers: How the Wider AI Stack Can Stop the Worst AI Misuse (MIT Press forthcoming 2027) ch 6.

Citation data for reference managers: BibTeX or RIS. This is a preprint of a book forthcoming from MIT Press in 2027. Page numbers follow this preprint and will differ from the published edition.