Stop Pretending AI Governance Is Brand New
AI governance feels like development déjà vu—big promises, fragile accountability, and a fight over who gets to define the rules.
I come from the world of international development—where promises are big, but outcomes depend on context, politics, and who’s in the room when decisions are made. I’ve always been drawn to the parts of the work that involve asking hard questions: What’s working, what’s not, and how do we know? Recently, as the development sector contracted, I shifted into social impact consulting—supporting organizations in building strategies that learn, adapt, and course-correct. And I'm paying close attention to artificial intelligence because: a) it’s not going away, and; b) it could be part of the solution to challenges we’ve been wrestling with in the social sector for decades.
AI governance today reminds me of that space I entered as an early practitioner—full of bold promises, existential anxieties, and intense debates about how to move forward responsibly. What stands out most is how quickly polarized narratives take hold. Discussions often collapse into oversimplified binaries, obscuring the layered concerns actually in play: ethical dilemmas, economic impacts, social equity, geopolitical tensions, environmental risks—and, too often, overly specific and prescriptive solutions.
Just like in development, many proposed strategies for governing AI are tightly bound to one domain or worldview—failing to account for the diversity of contexts in which these systems are being developed and deployed. Whether it’s assumptions about Western legal norms or one-size-fits-all technical fixes, too many approaches overlook the complexity of real-world environments.
There’s a common temptation for humans to frame AI governance in binary terms. On one end of the spectrum are the accelerationists. Accelerationists push for rapid innovation with minimal interference—trusting market forces and technological ingenuity to sort things out.
On the other are the doomers, those who are deeply concerned that advanced AI poses existential threats. Doomers call for sweeping regulation or even a pause on development.
These poles dominate the discourse—especially in think tank bubbles and on social media—but they don’t reflect the full range of views shaping real-world decisions. As with most things in development—and life, more generally—the reality is far more nuanced. Many researchers, practitioners, and policymakers fall between those extremes, trying to balance innovation with accountability, progress with precaution.
Some benevolent leaders are asking practical questions: How do we build AI systems that serve the public good? What mechanisms ensure transparency and fairness? How do we govern in ways that adapt as the technology evolves? As for the malevolent ones—well, we’re not entirely sure who they are and what they’re asking, and that uncertainty is part of the challenge. Then there are the wild cards like Elon Musk, who has used strong language warning that AI could become an “immortal dictator,” while he simultaneously builds a "maximally truth-seeking AI," and we’re supposed to trust him.
This messy reality aligns with my experience in international development, where billionaires are full of grand ideas for subjects in which they have no expertise, and durable outcomes rarely emerge from rigid ideologies. The best results come from iterative, evidence-informed approaches that acknowledge complexity and are inclusive, keeping the communities at the center. It’s the same principle echoed in the disability rights movement’s rallying cry: Nothing about us without us. The most effective and equitable solutions emerge when the people affected are directly involved in shaping them.
This post kicks off a four-part series on AI governance. Over the next three installments, I’ll explore how principles from development—like adaptive management, participatory evaluation, and continuous learning—can inform a more grounded, actionable approach to governing AI:
Part 1: Mapping the Landscape — Exploring the spectrum of AI governance perspectives, including accelerationists, doomers, and the many nuanced positions in between.
Part 2: What Development Can Teach Us — Highlighting transferable lessons from international development: how participatory, context-driven governance improves outcomes.
Part 3: Governing in Practice — Real-world tools and policy levers for responsible AI governance, including model audits, forecasting, regulatory sandboxes, and more.
AI is at a pivotal moment—rich with opportunity, but also carrying real risk.
Effective governance isn’t just about avoiding harm. It’s about shaping technology in ways that are safe, fair, and deeply human.
👋 Join the Series
If you're committed to helping guide AI’s future toward something more thoughtful and inclusive, I hope you’ll stick with me for the rest of this series.
This post is part of a series from Anthralytic, a strategy and evaluation consultancy helping mission-driven organizations make sense of complexity, measure what matters, and adapt in real time. If you're navigating the future of impact, we’d love to be in conversation.