When Society Asks Us to Do More

A Call to Action for Business

Next week is Climate Week here in New York, and I expect many CEOs to hit town. But it won’t just be business as usual. As the data center war wages and AI warnings become truly terrifying (we should pay attention when they use the word “existential”), I predict society will be asking business leaders to step up in ways we haven’t seen for a while.

Yes, we need stronger and smarter public policy. Let’s lobby for it! Let’s be sure to vote!

And yes, the leading AI companies could and should self-regulate.

But we can’t just wait around for these developments. Business is going to be called to action.

My data center playbook tells us what needs to be done in that space—and what should be avoided. And it should work for AI safety too.

Here’s a brief recap:

Don’t: Do things in secret using NDAs and hope communities don’t notice.

Don’t: Make commitments that rely solely on self-verification. No one will believe you.

Do: Put forth transparent, science-based plans for managing important outcomes for society— things like water, land, clean energy, utility bills, emissions, and, yes, AI safety.

Do: Work in real partnership with a credible NGO.

Why an NGO? Because the missing ingredient here is trust. Many companies are already making pretty decent commitments—but they’re self-declared and self-verified, so audiences discount them sharply. A respected independent group that can check the claims—and can walk away publicly if the company falls short—is a different thing entirely.

Would this be easy to do? No. It will be very difficult. The opposition has gotten very intense and politically charged. And we don’t know how many NGOs with the necessary science and the credibility would be willing to take on these challenges. But when I was the CEO of The Nature Conservancy, we often did work like this. We built partnerships with companies like Dow Chemical, Rio Tinto, JPMorgan, SwissRe, and Alibaba. The partnerships worked.

What’s happening with AI right now reminds me of what happened after Three Mile Island. The nuclear utilities built INPO— their own tough, independent inspectorate—because one more accident would have finished the industry for everybody. Something like that is needed for AI.

But here’s the problem: the AI labs may not be able to do this by themselves. OpenAI and Anthropic compete too intensely. The nuclear utilities could cooperate because they weren’t racing one another for the same exact opportunity.

But back to the hyperscalers—companies like Microsoft, Amazon, Meta, and Google. They could and should champion AI safety. Indeed, it should be part of the new and improved commitments we recommend they make around their data centers.

The hyperscalers are the AI labs’ compute providers, capital sources, customers, and—in some cases—large shareholders. They don’t own enough equity to give orders, but they have an awful lot of leverage and enormous standing. And they have every reason to use it. It’s the hyperscalers’ data centers that don’t get approved when the public is frightened. The labs’ license to develop more advanced AI and the hyperscalers’ license to build are part of the same social contract.

And there is plenty of room for other business leaders to step up. Wall Street firms—who are benefiting hugely from the AI boom—would be an obvious place to start

Of course we also need China to engage. We’ve seen them do so before. Again, business leaders in the US have shown in the recent past that they can help get such critical cooperation underway (even when government-to-government diplomacy moves slowly).

I’m just scratching the surface. These pressures will keep escalating. CEOs who look ahead thoughtfully on behalf of their stakeholders will have to prioritize addressing these challenges. Let’s get that dialogue underway now. We can start during Climate Week.

Onward,