As AI becomes integral to business, defining its governance is critical. Two models, provider sovereignty and deployer sovereignty, offer contrasting approaches. The debate is far from settled.
As artificial intelligence (AI) systems become increasingly embedded in organizational workflows, the question arises: who should have the final say over AI decisions? Two competing governance models offer different answers. On one side, frontier-provider sovereignty suggests the creators of the most advanced AI models should hold the reins. On the other, action-centered deployer sovereignty argues that those deploying AI, and bearing its consequences, should be in charge.
Governance Models in the Spotlight #
Frontier-provider sovereignty is gaining traction through frameworks that emphasize the responsibilities of AI model providers. This includes rigorous model testing, transparency obligations, and compute-related controls. Essentially, it places significant authority with those who develop advanced AI technologies.
Contrasting this is the action-centered deployer approach. Here, the focus is on the organizations that integrate AI into their processes, arguing they should have the final authority over high-impact actions. These organizations are the ones facing legal, operational, and commercial risks, making them logical candidates for decision-making power.
The Global Perspective #
Analyzing global governance frameworks reveals a surprising trend. The European Union's AI Act, the NIST AI Risk Management Framework in the U.S., and similar policies from Singapore, Japan, and Canada, all hint at a preference for distributed operational accountability. This suggests a tilt towards placing more control in the hands of those deploying AI, rather than solely on the frontier providers.
However, this isn't a straightforward consensus. Each region has its nuances, but the common thread is clear: the increasing complexity of AI within enterprises and the decline in transparency from providers necessitate a governance model that focuses on those implementing AI solutions.
A Layered Solution #
The paper's conclusion is anything but absolutist. While strong oversight is needed for frontier capabilities, the final authority over concrete enterprise actions should rest with those who implement and experience the outcomes. Is this the perfect balance? Perhaps. But as AI continues its march into the heart of business processes, this layered approach might be the most pragmatic.
So, who should hold the reins? While the debate rages on, one thing is clear: both providers and deployers have essential roles to play in the responsible governance of AI. As enterprises race to adopt AI, will they be equipped to handle its consequences, or will we see a power struggle in the boardrooms? It's a narrative that will shape the future of AI in our workplaces.
Get AI news in your inbox
Daily digest of what matters in AI.