The Illusion of Sovereign AI
As governments and private sector organizations worldwide scramble to integrate artificial intelligence, a cautionary voice has emerged from the architect of Britain’s own digital infrastructure. Mike Bracken, the founder of the Government Digital Service (GDS), argues that the current 'sovereign AI' narrative—focused heavily on building domestic data centers and local compute—misses the true danger: a creeping loss of organizational autonomy.
In his recent work, Digital Sovereignty: The Power to Decide, Bracken suggests that institutions are mistakenly equating geographical control of infrastructure with the ability to remain agile. He posits that sovereignty is not about where a server resides, but whether an organization can change its technology stack or pivot its strategy without being crippled by the weight of its previous vendor choices.
The Trap of Gradual Dependency
Bracken’s central thesis is that institutional independence is rarely surrendered in a single, catastrophic moment. Instead, it is eroded through a series of 'reasonable' decisions—choosing the most convenient AI model today, integrating a proprietary API tomorrow, and building workflows around a specific supplier's ecosystem the day after. Over time, these individual choices create a labyrinth of dependencies that make moving to an alternative virtually impossible.
The risk, according to Bracken, is that when an organization eventually needs to change course—due to shifting market demands, policy changes, or vendor price hikes—it will find itself 'locked in.' This effectively hands over the reins of power to a small circle of technology giants who control the foundational models upon which these critical systems now rely.
Why It Matters
- Strategic Flexibility: True sovereignty lies in the ability to swap suppliers or technologies without institutional collapse.
- The Infrastructure Fallacy: Domestic compute capacity does not guarantee independence if the underlying models and software remain proprietary and closed.
- Economic Resilience: Over-reliance on a few AI providers risks creating 'tech monopolies' that dictate the operations of both public sector bodies and large-scale enterprises.
- Long-term Decision Making: Leaders must evaluate AI adoption not just for immediate gains, but for the potential technical debt and loss of agency it imposes on the future.
Reframing the AI Conversation
The current push by governments to subsidize domestic hardware is a start, but Bracken argues it is incomplete. By focusing purely on 'sovereign' hardware, policymakers may be ignoring the software layer—the models and platforms—that actually dictates how an institution functions. If the intelligence layer is monolithic and opaque, the hardware it runs on is largely irrelevant to the goal of maintaining sovereign decision-making power.
For those navigating the current AI landscape, the takeaway is clear: before integrating external AI tools, organizations must consider whether they are expanding their future capabilities or quietly shrinking their window of choice. As Bracken notes, every piece of technology should either empower an institution or create a barrier; the danger today is that many organizations are choosing the latter, unaware that they are slowly giving away their right to decide for themselves.
