Home Institute The Institute Eva M. Erpenbach Research The Research Architecture Institute Essays Research Dialogues Guest Essays External Contributions Fellowship The Fellowship Fellows Fellowship Contributions Partners Project Partnerships Project Partners Young Research Contact Legal Notice
DEEN
← Institute Essays
Institute Essays / 01

Governance Silence Is Ambiguous

The absence of reported matters is evidence within the categories that produced it. It does not, on its own, establish that no decision-relevant condition existed.

Eva M. Erpenbach
Founder & Scientific Director, Erpenbach Institute

Corporate governance pays close attention to how boards receive information. Questions now include who prepares board materials and how management structures them. The same attention extends to what reaches the agenda and to the effect artificial intelligence may have on the information available to directors. The Artificially Intelligent Boardroom, by David F. Larcker, Amit Seru and Brian Tayan, published on the Harvard Law School Forum on Corporate Governance on April 8, 2025, describes current governance practice as placing directors in a responsive position to management. That position depends on what information management provides and on how that information is structured or framed. The authors consider how AI may change that relationship.

Those questions begin with something that already exists as information about a recognizable matter. An earlier question concerns how a condition becomes a matter at all. A condition may acquire institutional form as a risk or an exception. It may also become recognizable as an incident or a material change. In other cases, it becomes an issue capable of entering a decision process. That earlier layer is the subject of The Prior Order of the Decidable.

Before the Matter Exists for Governance

Governance can act only on objects that have become available to it in some institutional form. A risk cannot be assessed until the condition is recognizable as one, and an exception cannot be escalated until it has acquired that status. Review depends on the same prior step: the institution must first recognize that there is a decision to review.

The paper describes this preceding ordering as formation. It establishes conditions under which later decision processes become possible. Its difficult feature is that an effective prior ordering need not remain visible as a decision once it has taken effect. Institutional distinctions can emerge through earlier classifications and inherited boundaries. Accumulated practices and previous judgments can have the same effect. Later actors encounter their combined effect as part of the environment within which decisions are made.

The consequence is more important than the historical origin of any one boundary. A prior determination can remain effective after it has ceased to appear as a determination. Reconstructing when a threshold was adopted or how a category developed therefore does not by itself restore its status as an open decision. An institution may know the history of a distinction and still encounter that distinction in present governance as a feature of the world that is no longer open to present choice.

Boards consequently see the objects produced by the ordering, such as an identified risk or a reported incident. By the time an exception is classified or a change treated as material, the ordering has already done part of its work. The same is true of an item that reaches the agenda. The conditions that determined whether something could acquire one of those forms may no longer appear alongside it as a separate governance event. The issue therefore arises before questions of information quality. It concerns the point at which there is something institutionally constituted to inform the board about.

Framing Begins Later

Management framing matters because a recognized issue can be described in different ways. A presentation may emphasize one aspect while compressing another. Context can alter how the issue is understood; the choice of metric can do the same. Agenda control likewise presupposes that a matter is identifiable enough to be acted on. Once identifiable, it may be included or excluded. If included, it may receive priority or be deferred. Both forms of governance operate on an object that already exists for institutional purposes.

The prior-order question concerns the preceding transition. A developing condition may gradually become identifiable as a cyber incident or a conduct problem. It may instead acquire institutional form as model risk or as a material operational change. Once that form exists, ordinary governance questions become available. Who received the information? Was it escalated? The board can then examine whether the presentation was accurate and whether it asked the right questions.

Before institutional formation, those questions lack a stable object. This is why the issue cannot be reduced to a person withholding information or to a badly designed taxonomy. Some elements of a prior ordering may have identifiable authors. The structural point concerns what happens later, when their effects persist after their status as decisions has faded. Governance then operates within the result of earlier ordering while experiencing that result as the ordinary field in which current decisions occur.

A Rarely Visible Formation Event

The rapid emergence of AI risk provides an unusual contrast because part of this formation can be observed retrospectively. The Conference Board's From Principles to Practice: Governing AI in the Corporation, by Andrew Jones, published on April 22, 2026, reports that the share of S&P 500 companies disclosing AI as a risk rose from 12 percent in 2023 to 83 percent in 2025. These disclosure figures are based on S&P 500 disclosure data as of December 2025. The figures support a narrower conclusion about institutional formation. Over a short period, "AI risk" became a widely used disclosure category among S&P 500 companies. The figures leave open how much underlying AI risk changed during the same period and what individual boards knew.

AI is useful here precisely because the transition left an unusually visible trace. Historical disclosure data make the formation of the category observable. Prior-order effects are often harder to see because the resulting categories are already taken for granted. The same is true of the distinctions and thresholds that sustain them.

In 2025, asking whether a company discloses AI risk refers to a recognizable corporate category used across most of the S&P 500. Two years earlier, the absence of such disclosure at most companies carried a more ambiguous meaning. It could not, on its own, establish that no decision-relevant AI-related condition existed. The institutional form through which those conditions would be described was itself still becoming established.

Once a category stabilizes, its history tends to recede from ordinary governance reasoning. What was once a developing institutional distinction can begin to look like an obvious division in the world. The category remains visible; its status as the product of prior ordering becomes less so.

What Silence Can Establish

The practical implication is an inference rule concerning the evidentiary reach of governance silence. Larcker, Seru and Tayan note that a board may rely on information provided by management when no red flags are present. This raises the prior question of what had to happen before a condition could become recognizable as a red flag.

Governance routinely produces negative statements in different forms. A report may show no incident or exception, while separate records show no material change and no escalation. Elsewhere, governance may record that no risk was identified. Such statements can carry substantial information. Their evidentiary reach depends on what had to occur before something could acquire the institutional status being counted.

An absence of reported incidents establishes something about matters that became incidents for institutional purposes. It does not automatically establish that no underlying decision-relevant condition existed before the point at which it could acquire that status. The same logic applies to other governance categories.

A condition that never acquired the relevant institutional form was never a deficient entry in a category. It was never an entry.

That distinction matters because the visible result can be compatible with different underlying states. A quiet governance channel may reflect the genuine absence of relevant conditions. It may also reflect conditions that have not yet acquired the form through which the channel receives them. The report need not be inaccurate in either case. The question concerns the inference drawn from its silence.

A board can reasonably treat the absence of recognized matters as evidence within the governance structure that produced that result. A stronger proposition, that nothing decision-relevant existed outside the field of matters that became recognizable, requires an additional basis.

The ambiguity becomes more important during periods of change. New technologies can create unfamiliar dependencies before institutions settle the categories through which they become governable. Business models and organizational conduct can move through the same interval. During that interval, apparently quiet governance channels can coexist with meaningful change upstream of them.

Why Another Control Does Not Remove the Problem

A natural response would be to create an additional mechanism charged with examining how matters become visible to governance. Such a mechanism may improve governance in particular settings. It would nevertheless need boundaries for what it attends to and criteria for deciding when something becomes a matter for review. It therefore operates through a prior ordering of its own.

Boards should continue improving information flows and challenging management presentations. They should revise taxonomies where appropriate. Access to relevant information can also be widened and escalation strengthened where needed. These measures address important downstream problems and may materially improve what becomes visible. The structural observation here is narrower: adding another review layer does not remove prior ordering from governance.

The consequence for board practice is therefore deliberately limited. It concerns the confidence attached to an inference.

A board may rely on what its governance system brings before it while remaining disciplined about what the system's silence can establish. The absence of matters within a recognized governance category may support the conclusion that none entered that category under its existing conditions. It cannot, without further basis, establish that no decision-relevant condition existed before institutional formation.

Corporate governance necessarily works through formed objects. A board cannot deliberate over a risk until it has become recognizable as one for institutional purposes. Exceptions and other matters likewise have to acquire institutional form before they can enter governance. The prior order remains part of that process even when it no longer appears as a decision.

Governance silence is evidence. Its meaning depends on what had to happen before anything could become visible enough to be silent about.