AI's Next Leap: Who's Really in Control?

Jan de Vries · 2026-10-09

### The AI Governance Problem Nobody's Talking About Imagine you're building a house of cards. Each card relies on the one below it. Now imagine that with every new card you add, the entire structure gets more complex—and more fragile. That's basically where we are with AI governance, according to Vendan Ananda Kumararajah. He's been ringing alarm bells about something most of us overlook: the assumptions we bake into AI systems today could multiply into full-blown disasters tomorrow. And the scary part? We're not really set up to catch those mistakes before they cascade. ### Why Independent Oversight Isn't Just Nice—It's Necessary Here's the thing. When AI systems start designing their successors (and yes, that's already happening in some labs), they inherit all the flawed thinking of their predecessors. It's like a game of telephone where the message gets more garbled with each pass—except the stakes are global. Kumararajah argues that without independent oversight, we're essentially letting AI systems grade their own homework. That's a recipe for disaster. Independent oversight means having fresh eyes—human eyes—that can spot problems the original developers might have missed or chosen to ignore. But wait, isn't that what regulators are for? Not exactly. Most regulatory bodies move slower than molasses in January, and AI development is sprinting ahead. By the time a law passes, the technology has already evolved into something else entirely. > "The real danger isn't that AI will suddenly turn evil. It's that we'll keep building on shaky foundations until the whole thing collapses—and we won't see it coming." ### The Flawed Assumptions We Keep Making Let's get specific. What kind of flawed assumptions are we talking about? - **Assumption 1: More data equals better outcomes.** Not always. Biased data leads to biased decisions, and AI doesn't know the difference. - **Assumption 2: AI systems are objective.** They're only as objective as the humans who program them—and humans are anything but objective. - **Assumption 3: We'll have time to fix things later.** By the time we realize something's wrong, the system might be too entrenched to change. Each of these assumptions gets passed down to the next generation of AI. It's like compound interest—except instead of money, we're compounding risk. ### What Does Good Governance Actually Look Like? Kumararajah isn't just complaining. He's proposing a framework where independent experts—ethicists, technologists, and yes, even philosophers—have a real say in how AI systems are developed and deployed. Think of it like having a building inspector for AI. You wouldn't build a skyscraper without someone checking the foundation, right? Same principle applies here. This isn't about slowing down innovation. It's about making sure the innovation doesn't collapse under its own weight. Independent oversight can actually speed things up by catching problems early, before they become expensive—or impossible—to fix. ### The Clock Is Ticking We're at a crossroads. AI capabilities are doubling faster than our ability to govern them. If we don't rethink our approach now, we risk an intelligence explosion we can't control. That's not science fiction. That's just math. And the solution isn't rocket science either—it's common sense. Independent oversight, diverse perspectives, and a willingness to question our own assumptions. The question is: will we act before it's too late? Or will we keep building that house of cards, hoping it won't fall?