Piece 8

A recent parliamentary warning in the United Kingdom has brought renewed attention to a structural question that extends well beyond the domain of artificial intelligence itself: the capacity of institutions to govern decision-making processes that are no longer fully human.

With a significant share of the financial sector now relying on artificial intelligence, the use of automated systems is no longer peripheral. It has become embedded within the architecture of decision-making.

This transformation, however, has not been matched by an equivalent evolution in governance frameworks. Uncertainty around accountability, limited supervisory clarity, and the opacity of automated processes reveal a structural gap: decision-making is being delegated without a corresponding redefinition of authority. This is precisely where the issue acquires institutional significance.

Artificial intelligence does not merely introduce efficiency gains. It alters the way decisions are produced, distributed and justified.

As multiple institutions increasingly rely on similar models, trained on comparable datasets and shaped by converging logics, a distinct systemic risk begins to emerge. The concern is not only individual error, but the possibility of “herd behaviour”: decisions that are formally independent, yet structurally aligned. In moments of stress, such alignment may amplify shocks rather than absorb them.

In this context, stability becomes less a function of individual institutional robustness and more a function of diversity, autonomy and intelligibility within decision-making processes.

The deeper challenge, therefore, does not lie simply in the possibility of error, nor only in the technical refinement of these systems. It lies in whether institutions can preserve effective control over decision processes whose internal logic becomes progressively less visible.

At the point where decision-making no longer has a clearly identifiable author, the challenge ceases to be purely technical, it becomes institutional. Because when authority becomes diffuse while the consequences of decisions remain fully tangible, the demand for legitimacy does not diminish. It becomes more exacting.

For policymakers, regulators and leaders operating in complex systems, the central question is no longer whether technology can support decision-making.

It is whether institutions retain sufficient capacity to govern that delegation without dissolving accountability, obscuring judgement, or weakening the legitimacy upon which their authority ultimately depends.

—

The Guardian (2026)
UK exposed to “serious harm” by failure to tackle AI risks, MPs warn