Reversibility Standards for Public Digital Infrastructure
Turning reversibility from a principle into an assessable procurement criterion, so accepting an irreversible commitment becomes a recorded decision rather than a…
Our programmes convene the people who have to make ethical infrastructure real — regulators, engineers, clinicians, educators and civil society — around problems none of them can solve alone. Seven categories of work are set out below. Where a programme is still being convened we say so, rather than describing intentions as achievements.
Each is a distinct kind of commitment, with its own criteria for what counts as done.
Standing groups convened around a single governance question, bringing regulators, engineers and affected practitioners into the same drafting process. A working group has a defined question, named participants and a date by which it must produce something. Groups that cannot state all three are not yet working groups.
Enquire about participating →Time-bounded empirical and analytical work producing publications, datasets and methods. We are particularly interested in whether governance mechanisms change outcomes in practice — a question asked far less often than whether they exist on paper.
Propose a collaboration →Curriculum development and teaching material for schools, universities and professional training bodies, aimed at building ethical reasoning into technical education rather than appending it as a final-year module. Our teaching material is released for reuse and adaptation.
Discuss teaching →Small, instrumented trials of a governance mechanism inside a real institution, with agreed measures and a published account of the result — including when the mechanism does not work. A recommendation that has never been tried in an institution is a hypothesis.
Host a pilot →Open calls inviting researchers, engineers and institutions anywhere to work on a defined problem in parallel, with published methods so results can be compared rather than merely collected. Useful where the problem is shared but the institutional context is not.
Register interest →Formal submissions to consultations, legislative committees and standards bodies, drafted to be usable by the people who must implement them. We publish our submissions in full, including where our advice was not taken.
Read our submissions →Long-term commitments with institutions that embed our review methods into their own processes. Partnership means shared work and mutual obligation — it is not an endorsement in either direction, and we say so publicly.
Propose a partnership →
Turning reversibility from a principle into an assessable procurement criterion, so accepting an irreversible commitment becomes a recorded decision rather than a…
5 August 2026
Shared technical capacity across supervisory authorities that individually cannot justify the headcount but together comfortably can.
5 August 2026
Five contract clauses — audit access, performance disclosure, override preservation, exit, and change notification — drafted to be inserted directly into public…
5 August 2026
An open call inviting teams anywhere to measure human override of automated decisions using a shared published protocol, so results can be…
5 August 2026
A controlled trial of whether removing throughput pressure restores meaningful human override in a deployed decision-support system — with the result published…
5 August 2026
Assignment material that embeds ethical reasoning inside technical coursework, released for reuse so the marginal cost to a lecturer is close to…
5 August 2026
A working group drafting a standard that its implementers did not shape produces a document nobody implements. We convene rather than pronounce, which means the programmes above depend on practitioners joining them.
For regulators, engineers and caseworkers who will have to apply what we draft.
EnquireCo-authored work, shared methods, or peer review of our working drafts.
ProposeTest a governance mechanism inside your institution, with the result published either way.
Discuss