Our Vision

Every generation inherits infrastructure it did not choose. Roads, courts, curricula, clinical protocols, payment systems, electrical grids — these are not neutral conveniences. They are decisions made durable. They determine who is heard, who is counted, who is treated, and who is left to argue with a system that cannot hear the argument.

What is new is the speed. A technology can now be conceived, deployed and embedded in the daily life of hundreds of millions of people before any deliberative body has finished forming a view on it. The institutions we rely on to weigh consequences — parliaments, regulators, professional bodies, courts, universities — were designed for a pace of change that no longer exists. They are not failing through negligence. They are being outrun.

This gap between capability and governance is the defining condition of our moment. It is not a temporary lag that will close on its own. Left alone it widens, because capability compounds and deliberation does not.

Ethical Infrastructure exists to help close that gap — not by slowing invention, but by building the institutional capacity to keep pace with it.

Most ethical failure in technology is not the work of bad actors. It is the accumulated result of ordinary decisions made by people who had no framework for the question in front of them, no authority to pause, and no obligation to consider anyone not in the room. A procurement officer selecting a triage system. An engineer choosing a default. A ministry adopting a standard drafted elsewhere. Each decision is small. Together they constitute the moral architecture of a society, assembled without anyone having designed it.

We work on the layer beneath the headlines: the standards, review structures, procurement rules, professional norms, accountability mechanisms and institutional habits that determine what gets built and what gets stopped. That layer is unglamorous and it is decisive. It is also, unlike public opinion, something that can be deliberately improved.

Our purpose is to help ensure that innovation remains aligned with human dignity, responsibility and the common good — and to make that alignment structural rather than dependent on the conscience of whoever happens to be in charge.

We do not treat these as separate problems requiring separate specialists. They are the same problem appearing in different institutions, and the solutions transfer between them more often than disciplinary boundaries suggest.

Decisions without accountability. Automated systems increasingly make or shape consequential decisions about people — in credit, employment, welfare, immigration, policing, healthcare and education — while diffusing responsibility so widely that no one can be held to answer for the outcome.

Capability concentrated, consequence distributed. The capacity to build the most consequential systems sits with a small number of organisations. The consequences are borne by everyone. Institutions designed to represent the many have limited purchase on decisions made by the few.

Regulatory fragmentation. Infrastructure crosses borders; regulation does not. Fragmented rules produce forum-shopping, incoherent obligations for those trying to comply, and a floor of protection set by whichever jurisdiction cares least.

Erosion of the shared record. Democratic deliberation requires a broadly shared account of what is true. Systems optimised for attention rather than accuracy corrode that foundation, and with it the possibility of resolving disagreement by argument rather than force.

Compounding environmental cost. Digital infrastructure has a material footprint — energy, water, minerals, land. It is routinely treated as immaterial in the decisions that authorise it.

Institutional capacity. Many of the bodies charged with oversight lack the technical staff, the statutory powers, or the time to exercise it meaningfully. Good rules administered by an under-resourced regulator are not protection.

Unrepresented interests. The people most affected by infrastructural decisions — including those not yet born — are structurally absent from the rooms where those decisions are made.

We do not mean a code of conduct, a set of principles, or a statement of values. Those exist in abundance and have proved easy to sign and easy to ignore.

By ethical infrastructure we mean the load-bearing structures that make ethical outcomes the default rather than the exception: standards with technical specificity, review processes with real authority to stop work, procurement conditions with consequences attached, professional obligations that survive commercial pressure, audit rights that are exercised, and appeal routes that a person without a lawyer can actually use.

The test we apply is simple. If acting well depends on individual courage, the infrastructure has failed. Courage is admirable and it is not a system. Our aim is to make doing the right thing the path of least resistance — and to make doing the wrong thing structurally difficult, visible, and answerable.

Our method follows from our subject. Infrastructure is built by many hands over long periods, and it cannot be improved from outside the practice.

We work upstream. Ethical review applied after deployment is damage assessment. We work at specification, procurement and standard-setting, where a decision still costs little to change.

We convene rather than pronounce. The people who must implement a standard should have shaped it. Our working groups bring regulators, engineers, clinicians, educators and affected communities into the same process.

We publish in usable form. A recommendation a regulator cannot draft from, or an engineer cannot implement, is an opinion. Our output is intended to be operational.

We publish openly. Our public research is free to read, download and cite, without registration. Work intended to inform public decisions should not sit behind a form.

We disclose what shapes our judgement. An institution that examines the incentives of others must be legible about its own.

We revise in public. Working on fast-moving systems means being wrong sometimes. We would rather correct a published position than defend one.

Infrastructure outlives the arguments that produced it. A standard adopted this decade will govern systems still running when the people who wrote it have retired. A default set once is inherited by everyone who comes after, usually without knowing a choice was ever made.

This gives our subject an unusual moral structure. The people with the strongest interest in these decisions are the people least able to participate in them. They cannot lobby, vote, litigate or object. They will simply live in what we build.

We treat that absence as a design problem rather than a sentiment. It means asking, of any proposal, what it forecloses; preferring reversible choices over efficient irreversible ones; declining to trade a permanent constraint for a temporary gain; and being explicit when a decision transfers cost to people who cannot answer back.

We commit to leaving those who follow us more room to decide than we were given — better instruments, clearer records, fewer locked doors, and a functioning capacity to change their minds.

We are building an institution intended to last longer than any technology it examines, and to be useful across the disagreements it will inevitably outlive.

The world we are working towards is not one in which technology is constrained, but one in which its direction is genuinely chosen: where the ethical review of consequential systems is as routine and as unremarkable as their security review; where a person subject to an automated decision can obtain a reasoned explanation and a human reconsideration as a matter of course; where standards hold across borders because the institutions that set them cooperate; and where the interests of those not present are represented as a matter of structure rather than goodwill.

None of this is achieved by one organisation, and none of it is achieved quickly. It is achieved by many institutions making incrementally better decisions, sustained over decades. Our work is to make those decisions easier to make well.

This is an open project. It requires researchers, practitioners, regulators, funders and critics — particularly critics. If you have read this far, there is likely a version of this work that needs you.