Briefing: Institutional capacity in technology supervision
An eight-page briefing for legislators on why enforcement is limited by staffing rather than statutory powers.
Democratic government depends on conditions that technology can either support or quietly dissolve: a broadly shared account of what is true, deliberation capable of changing minds, and administration that citizens can understand and contest. Attention usually falls on the electoral layer — campaigning, disinformation, platform influence — and that attention is warranted. But the state's relationship with most citizens is administrative rather than electoral, and it is mediated by systems.
A person encounters the state when a benefit is calculated, a visa is assessed, a tax return is flagged, a child's school place is allocated, or a police resource is directed towards their street. Each of those is increasingly shaped by an automated process. Where that process cannot be explained, cannot be contested, and cannot be traced to anyone accountable, the citizen's relationship with the state changes character — from a relationship with an institution that owes them reasons to an encounter with a system that produces outcomes. That shift happens without any vote being taken on it.
We work on the administrative layer as seriously as the electoral one. Transparency obligations for public-sector systems that go beyond publishing a register. Procurement conditions that preserve accountability rather than contracting it away. Appeal routes a person can navigate without a lawyer, terminating in someone empowered to reverse the decision rather than in a review of whether the process was followed. And a practical requirement we think is close to foundational: that any decision significantly affecting a person's life remain subject to human reconsideration, as a matter of course rather than on request.
An eight-page briefing for legislators on why enforcement is limited by staffing rather than statutory powers.
An empirical study of what actually limits enforcement in technology supervision — comparing statutory powers against the staff, skills and time available…
Five contract clauses — audit access, performance disclosure, override preservation, exit, and change notification — drafted to be inserted directly into public…
Developing an auditable standard for the review of automated decisions affecting individuals, drafted with the regulators and engineers who would have to…
An open lecture on why the gap between capability and governance widens rather than closes, and what closing it would actually involve.
Most audit obligations in current legislation can be satisfied without anyone examining whether a system produces defensible decisions. This paper sets out…
The group is now accepting participants from outside the institute — particularly people who expect the draft standard to fail.
Current and former supervisory officials on what actually limits enforcement. The published note from this session shaped our capacity research project.
Every account of obligation we have runs through consent, participation or reciprocity. None of them reaches the people most affected by infrastructure…
We have published our full consultation response as submitted, including the proposal we opposed.
A conversation with a former supervisory official on what actually limits enforcement: not the absence of powers, but the absence of anyone…
Our full response to a public consultation, published as submitted. We supported three proposals, opposed one, and identified an omission we consider…