An AI Expert Warning: 6 People Are (Quietly) Deciding Humanity's Future

There is a question that does not get asked often enough in boardrooms, leadership teams, or strategy sessions: Who is actually making the decisions about how AI develops?

Not who is building the tools. Not who is running the demos. Who is making the foundational choices — about safety, about deployment, about what gets prioritised and what gets suppressed — that will shape AI for the next decade and beyond?

The answer, if you follow it honestly, is uncomfortable. It is a very small number of people. And most of the world is not paying attention to what they are deciding.

The illusion of distributed progress

AI feels broadly distributed. There are hundreds of companies building models, thousands of startups applying them, millions of people using the outputs. The ecosystem looks wide and open.

But that surface diversity hides a more concentrated reality. The foundational models that underpin most of what we call AI — the large language models, the multimodal systems, the frontier research — are being built by a very small number of organisations. And within those organisations, the critical decisions about capability, safety, deployment pace, and governance are being made by an even smaller number of people.

This is not a conspiracy. It is a structural consequence of how AI development has evolved. Frontier AI requires extraordinary compute, talent concentration, and capital. That combination naturally consolidates at the top.

"The people deciding the pace of AI development are not elected, not publicly accountable, and operating under competitive pressure that pushes toward speed."

Why this matters more than most leaders realise

The decisions being made at the frontier are not just technical choices. They are choices about values, risk tolerance, and what trade-offs are acceptable. They are choices about whether safety research keeps pace with capability research. About whether deployment happens when systems are ready or when competitive pressure demands it. About who gets access, on what terms, and under what conditions.

These are fundamentally governance questions. And they are being answered inside private companies, by small leadership teams, under commercial incentives that do not automatically align with broader societal interests.

That does not make the people involved malicious. Many are genuinely thoughtful about the risks. But the structure itself — concentrated, fast-moving, commercially driven, insufficiently regulated — creates conditions where important decisions get made without adequate external input, accountability, or oversight.

What the warning signs look like in practice

You can see the consequences of this concentration in several places.

You see it in the speed of deployment. Systems that experts flag as insufficiently tested get released because the competitive dynamics demand it. The window between development and deployment has compressed dramatically. What used to take years of evaluation now takes months — or weeks.

You see it in the framing of safety. Safety teams at frontier labs do important work. But they operate inside organisations where the primary incentive is capability advancement. When safety conclusions slow down releases, the institutional pressure is almost always toward finding a way to move anyway.

You see it in the governance vacuum. Meaningful AI regulation is still years away in most jurisdictions. The EU AI Act is the furthest along, but its scope is limited and its enforcement uncertain. In the US, executive orders have come and gone. In the Gulf, AI governance frameworks are emerging but not yet mature. In the interim, the companies building the most powerful systems are largely self-governing.

"When the people building the most powerful technology in human history are also the ones setting the rules for how it gets used, that is a governance problem — regardless of their intentions."

What leaders should actually do with this

This is not an argument for paralysis or for abandoning AI adoption. The technology is genuinely transformative and the business case for engagement is real.

But it is an argument for leaders to stop treating AI purely as a technology question and start engaging with it as a governance and risk question too.

That means understanding which systems your organisation depends on, who controls them, and what happens if the underlying conditions change. It means having views on AI policy — not just AI tools. It means contributing to industry standards conversations rather than waiting for others to set the framework.

And it means being honest with your teams and boards that the environment in which AI is being developed is not as stable, accountable, or well-governed as the marketing materials suggest.

The opportunity in the warning

There is also a real opportunity here for organisations that engage seriously. As regulation comes — and it will come — the companies that have thought carefully about governance, built responsible AI frameworks, and established accountability structures will be better positioned than those that treated it as an afterthought.

The organisations that understand AI as infrastructure — subject to the same strategic, risk, and governance thinking as any other critical system — will lead the next phase. The ones that treated it purely as a productivity tool will find themselves scrambling to catch up when the rules change.

The warning from AI experts is real. But it is also an invitation. The people deciding humanity's future do not have to be six people in San Francisco. They can include every leader who decides to engage seriously, think carefully, and show up to the governance conversation rather than leaving it to someone else.

---

---