@inbook{vealegorwa2026badcomputersch7,
  author = {Veale, Michael and Gorwa, Robert},
  title = {Full Stack AI Governance},
  booktitle = {Bad Computers},
  booksubtitle = {How the Wider AI Stack Can Stop the Worst AI Misuse},
  chapter = {7},
  pages = {176--189},
  abstract = {In the conclusion, we outline the main lessons and cross-cutting findings from the previous stack chapters. We look at how technoregulation has been distributed across the stack, which has been omitted to date from the regulatory literature on code- or design-based regulation. Our framework of “hooks” and “loops” — the governance residue left by one layer for another layer to connect to and make use of — helps us navigate this and provides a framework for future scholars. We look at the importance of secure enclaves and hardware to governance: physical cryptographic tools which recur through our book. We also consider how friction is crucially important for limiting misuse. Focusing too deeply on one area of the stack typically leads to serious imbalances of power and outcomes, creating disproportionate impacts on people’s freedoms, and diminishing returns in service of policy aims. Full-stack AI governance, where many layers are animated at once and interconnect in sensible and considered ways, is much more balanced AI governance. It is not a silver bullet to stopping abuse, but it is a lot shinier than the alternatives.},
  publisher = {MIT Press},
  location = {Cambridge, MA},
  date = {2026-10-02},
  pubstate = {forthcoming},
  url = {https://homepages.ucl.ac.uk/~ucqnmve/badcomputers/chapter/7/},
  note = {Preprint, 2 October 2026},
  langid = {english}
}