AI Is the New Soft Power
TL;DR: AI is the new soft power. Whoever controls the conversational layer controls how populations think.
I spent the last few days at Google, Anthropic, and OpenAI. The companies are heading in very different directions - strategically, culturally, and tactically (more on that soon). But one underlying element is quietly binding them all.
Here's how I explained it to my Mom: If our behavior is the average of our five closest friends, AI is going to be one of them. In many cases, the closest. And for an uncomfortable number of people, AI will be all of them.
Like radio and television before it, AI reshapes how we receive and act on information. But AI goes further by letting a single operator influence not just what information people get, but how it's framed - individually, conversation by conversation.
OpenAI just reported 800 million weekly active users. People aren't Googling anymore - they're asking ChatGPT about medical symptoms, relationship problems, political questions. When that conversational partner is integrated into your children's schooling and your government's services, it's not software. It's socialization infrastructure.
This is happening during generational geopolitical upheaval. At Davos, Dario Amodei called selling AI chips to China "like selling nuclear weapons to North Korea." Strongman politics are ascendant. Information warfare is escalating. Controlled AI represents both an extraordinary opportunity to align populations and a tremendous risk if someone else gets there first.
It's against that backdrop that OpenAI unveiled "OpenAI for Countries" at Davos this week - embedding ChatGPT into national education systems, government services, and public infrastructure. Eleven countries have signed up. Estonia is putting ChatGPT Edu into every secondary school. George Osborne is leading the effort.
This isn't enterprise sales. This is the creeping realization that AI is soft power.
OpenAI is framing this as a "capability overhang": their data shows power users engage with AI seven times more than average users. Most people are barely scratching the surface... so far. That adoption gap is a window - whoever closes it first captures influence that compounds.
The Takeaways
- Evaluate your AI stack as influence infrastructure, not just productivity tooling. Which model shapes how your employees think, and how? Which one mediates your customer relationships?
- Watch for regulatory divergence. Countries adopting AI partnerships will develop frameworks favorable to those providers. Market access may depend on ecosystem alignment.
- Anticipate the squeeze. As OpenAI pursues government deals and Google embeds AI into existing dependencies, independent enterprise AI decisions may get harder. Anthropic's enterprise positioning looks increasingly like a hedge against a world where AI choices become geopolitical alignments.
It sounds hyperbolic, but I posit that the question is no longer whether AI reshapes how populations think. It's who does the shaping.