AI operates through decisions made across many institutions. One organization builds a model, another adapts it, and another grants it authority to act. Each holds different information and controls a different part of its use. The people affected may have chosen none of them.
Effective oversight therefore depends on accountability reaching every organization that builds, adapts, or puts AI to work. Each must identify the decisions it controls, the authority it grants, and who answers for the consequences. Within the enterprise, a governing body holds oversight responsibility and a named officer answers for the program.
This is a shared public need. A failure in one institution can affect customers, patients, employees, counterparties, and confidence in institutions beyond its own boundaries. Warnings must reach someone with authority to act. People affected must have a route to raise concerns and challenge consequential decisions. Evidence from experience must inform the institutions and professions responsible for oversight.
Distributed Accountability makes that shared need concrete through specific responsibilities. It allows organizations to take informed risks while preserving a record of who decided, on what basis, and who can intervene. The program is the governance.
Responsibility across institutions
Builders and Buyers hold different decisions. A model developer decides what it releases and what it represents about its capabilities. An organization adapting a system decides what changes it makes and the uses it supports. An enterprise deploying that system decides where it may act, within what limits, and under whose authority.
Those decisions require distinct accountability. A supplier's assurance does not establish the enterprise's decision to use a system in a particular setting. An enterprise's authorization does not erase the supplier's responsibility for its own representations and decisions.
Knowledge must reach authority
An oversight program must identify who receives relevant warnings from suppliers, employees, customers, and affected communities, and who can decide whether a system should continue, be constrained, or be withdrawn. A reporting pathway is useful only when it reaches a party with defined authority.
The Center's role
The Center develops the shared language and reference materials that make these responsibilities intelligible to boards, executives, regulators, legislators, and the professions. Research, education, and convening allow institutions to learn from one another while retaining responsibility for their own decisions.