Blab AI
Who Holds the Stop-Button? The Missing Authority in Enterprise AI Governance
Adobe’s chief AI and data governance officer says the true test of an enterprise AI program isn’t the dashboards it produces, but whether anyone inside the organization can actually halt a model that is causing harm.
In a recent internal briefing, Adobe’s governance lead pointed out that while most Fortune 500 firms claim to govern their AI, few can answer who has the authority to shut a model down. The issue isn’t about notification or incident reporting; it’s about who can walk into a meeting and say, “We are stopping this.”
Adobe’s own approach illustrates the gap. The company has built a federated governance framework that assigns named owners to every AI system and establishes a centralized steering committee with escalation authority. That committee reports to the trust and security organization, not to product teams, giving governance a line of authority that is independent of the groups that ship AI products.
The urgency of this matter has been sharpened by the European Union’s Artificial Intelligence Act, which took effect on 1 August 2024. The Act obliges providers of high‑risk AI systems to keep detailed records and demonstrate meaningful governance, including documented decision‑making, clear lines of accountability, and the ability to show regulators who made consequential choices about an AI system and why.
Many large enterprises deploy hundreds of AI models—across customer service, hiring, content moderation, pricing, and fraud detection—under rapid, pressure‑driven timelines. Governance is often treated as a future task, resulting in a collection of tools that offer visibility but no mechanism for decisive action.
Adobe’s experience points to the missing piece: a dedicated “AI governor” with real authority. In practice, this role must be able to override product‑development incentives and stand ready to act when its decisions conflict with a roadmap. The governance function also needs a direct line to senior leadership, independent of the product teams that ship AI.
Because the EU AI Act places accountability at its core, regulators will look beyond policy documents or risk registries. They will demand a named individual or a clear governance body that can be held responsible for stopping a model.
For enterprises, the next step is to move beyond tooling and embed human accountability into the organizational design. That means appointing an AI governor, establishing a federated governance structure, and ensuring that the governance body can act decisively.
In short, the industry’s “governance theater” will end when companies can answer the simple question: Who in the organization can say “no” to an AI deployment and have that decision carry weight?
In a recent internal briefing, Adobe’s governance lead pointed out that while most Fortune 500 firms claim to govern their AI, few can answer who has the authority to shut a model down. The issue isn’t about notification or incident reporting; it’s about who can walk into a meeting and say, “We are stopping this.”
Adobe’s own approach illustrates the gap. The company has built a federated governance framework that assigns named owners to every AI system and establishes a centralized steering committee with escalation authority. That committee reports to the trust and security organization, not to product teams, giving governance a line of authority that is independent of the groups that ship AI products.
The urgency of this matter has been sharpened by the European Union’s Artificial Intelligence Act, which took effect on 1 August 2024. The Act obliges providers of high‑risk AI systems to keep detailed records and demonstrate meaningful governance, including documented decision‑making, clear lines of accountability, and the ability to show regulators who made consequential choices about an AI system and why.
Many large enterprises deploy hundreds of AI models—across customer service, hiring, content moderation, pricing, and fraud detection—under rapid, pressure‑driven timelines. Governance is often treated as a future task, resulting in a collection of tools that offer visibility but no mechanism for decisive action.
Adobe’s experience points to the missing piece: a dedicated “AI governor” with real authority. In practice, this role must be able to override product‑development incentives and stand ready to act when its decisions conflict with a roadmap. The governance function also needs a direct line to senior leadership, independent of the product teams that ship AI.
Because the EU AI Act places accountability at its core, regulators will look beyond policy documents or risk registries. They will demand a named individual or a clear governance body that can be held responsible for stopping a model.
For enterprises, the next step is to move beyond tooling and embed human accountability into the organizational design. That means appointing an AI governor, establishing a federated governance structure, and ensuring that the governance body can act decisively.
In short, the industry’s “governance theater” will end when companies can answer the simple question: Who in the organization can say “no” to an AI deployment and have that decision carry weight?
