Natasha Crampton says the definition of AI sovereignty has gotten muddy lately. She’s Microsoft’s chief responsible AI officer and she thinks most people are looking at this backwards. The real danger isn’t just keeping technology in or out of a country. It is the widening gap between the Global North and the Global South.
“We cannot let the digital divide become an甚至 greater AI divide,” Crampton told Euronews Next at the UN AI for Good Summit in last week’s Geneva proceedings.
The political landscape shifted fast. Remember when the Trump administration pressured Anthropic to bar non-US citizens from accessing Mythos and Fable? The move was partially reversed but the sting remained. Tech conferences have since become echo chambers shouting about “sovereignty” as if it meant building a digital iron curtain.
Crampton disagrees with that isolationist take. She argues AI sovereignty is actually about integration with local flavor.
Why AI sovereignty needs local values and global tools
It isn’t about rejecting global technology to build a inferior local clone. It’s about ensuring that when those massive global models land, they respect local culture. They need to understand norms. Idioms matter. A system that doesn’t grasp the nuance of your specific language fails its users.
That is where projects like LINGUA come in. Originally a European effort, it has expanded to Africa. This is a joint operation. Microsoft’s AI for Good lab partnered with the Gates Foundation, Google.org, and the Masakhane African Language Hub.
The goal is straightforward: collect real language data. Not scraped from the web and filtered through American biases, but ground-up local speech. This gives communities technical autonomy. It ensures they aren’t just consumers of foreign algorithms but participants in shaping them.
Connectivity is key here. But not just between computers. We are talking about connective tissue between governance bodies. The UN launched its first Global Dialogue on AI governance in early July. Why? To make sure no single nation writes the rulebook in isolation. Crampton wants these mechanisms to stop talking past each other. The next year should define unique roles for every player so progress doesn’t become repetitive work.
How to protect humanitarian workers in AI-driven conflict zones
Consider the “digital emblem.” This is a partnership between Microsoft, the Red Cross, and the International Telecommunication Union.
Cyberattacks are no longer abstract threats. Hospitals use cloud systems for patient records. Aid workers rely on logistics platforms. Attackers are targeting that infrastructure. The emblem is supposed to be a legal shield.
But a shield only works if people respect it. Microsoft wants governments to write the emblem into policy. They want tech companies to build recognition for that shield directly into their tools. And they want medical organizations to define the operational reality so the tech matches the field needs. It is about building trust into the code itself.
Is Microsoft actually a good digital neighbor?
People hate the sprawl of AI infrastructure. Water usage. Sky-high electricity bills. Local economies strained by data center expansion.
Crampton admits the footprint is heavy. But Microsoft is pivoting away from the standard “give us tax breaks” corporate playbook.
“We want to be good neighbours… so we have been taking steps actually to offer a community-first set,” Crampton noted.
Instead of draining the local treasury via tax exemptions, Microsoft aims to expand the tax base to fund schools and roads. On the resource side, they are using closed-loop cooling to manage heat. They are actively working to ensure that powering a server doesn’t spike your home electricity bill or dry up your local water supply.
This matters. The environmental cost of AI is tangible. It affects the air and water people touch every day. Ignoring that creates backlash that stumps the industry.
How can AI regulation adapt without becoming obsolete?
Europe is trying to get ahead of this curve. The European AI Office is talking to safety institutes in the UK, Canada, and the US. Cross-border coordination is necessary because testing science changes faster than bureaucracy usually allows.
Crampton urges humility here. Rules written three years ago might be garbage today if the underlying technology has leapt forward. Regulators often lag. There is a disconnect between what society expects and what the law allows.
“We need regulatory regimes to adapt… and ideally reduce the lag,” she said.
It is hard to legislate when you don’t fully understand the risk. Yet we have to try. International connectivity helps signal when an approach needs maturation. When new techniques emerge, regulations need to shift with them rather than breaking under the weight.
Who gets left out if we get this wrong? Probably everyone in the Global South waiting for a future that looks like their present. Not just a mirrored version of Silicon Valley.
