Who Controls AI?

Ask an AI assistant a question and something remarkable happens.

Within seconds, you can receive an explanation of quantum physics, help planning a vacation, feedback on a résumé, a bedtime story for your child, or a first draft of a business proposal.

It can feel almost effortless.

But behind that simple conversation sits an enormous system most of us rarely see.

There are companies building the AI models. Chipmakers manufacturing the processors needed to run them. Cloud companies operating massive data centers. Energy providers supplying electricity. Investors committing billions of dollars. Governments deciding what rules should apply. And developers turning those underlying systems into the applications we use every day.

In the first two parts of our AI for the Rest of Us series, we looked at what artificial intelligence can do and how we should decide when to trust it.

Now we arrive at a different question: Who controls AI?

The answer matters because AI is rapidly becoming more than another piece of software. If it becomes an important way we search for information, perform our jobs, educate our children, make purchases, and interact with the digital world, then the people and institutions controlling those systems will have enormous influence over everyday life.

And the answer turns out to be more complicated than simply naming a handful of technology companies.

The AI You Use Has an Owner

When you open an AI assistant, the experience can feel strangely neutral.

There's a box. You type a question. An answer appears.

But the system behind that box isn't neutral infrastructure that simply appeared on the internet. Someone built it.

Companies such as OpenAI, Google, Anthropic, Meta, xAI, Microsoft, and others are competing to develop increasingly capable artificial intelligence systems. They make decisions about how their models are trained, how people can access them, what features they offer, what they cost, and what safeguards surround their use.

Those decisions influence the AI experience you receive.

And increasingly, the companies building the models are intertwined with other technology giants providing the enormous computing resources AI requires. Microsoft remains a major shareholder and cloud partner of OpenAI, for example, while Amazon is Anthropic's primary cloud and training partner.

Understanding AI therefore requires looking beyond the chatbot itself.

Why Building AI Favors Giants

One reason control has become such an important question is simple: Building cutting-edge AI is extraordinarily expensive.

Training and operating today's most advanced models requires specialized computer chips, enormous data centers, sophisticated networking equipment, large amounts of electricity, highly skilled researchers, and access to enormous computing resources.

The scale can be difficult to comprehend. Anthropic announced in April that it had reached an agreement with Amazon to secure as much as five gigawatts of additional computing capacity over time, with Anthropic committing more than $100 billion to AWS technologies over ten years.

That is very different from the early days of the consumer internet, when a few talented programmers could build an important new company from a garage.

Startups can certainly still create innovative AI products. But developing the most powerful underlying models requires resources available to relatively few organizations.

That naturally creates concentration.

The companies with access to the most capital, computing power, chips, data centers, and talent have an enormous advantage.

Meet the AI Power Centers

So who actually controls AI? There isn't one answer.

AI developers such as OpenAI, Anthropic, Google, Meta, and xAI build the models and products consumers increasingly recognize.

But beneath them sits another layer.

Companies such as NVIDIA provide many of the specialized chips that make modern AI possible. Cloud providers including Microsoft Azure, Amazon Web Services, and Google Cloud supply enormous amounts of computing infrastructure.

Then there are companies building data centers, producing energy, developing applications, and integrating AI into everything from smartphones to office software.

The result looks less like a single industry and more like a stack:

Energy and chips

↓

Data centers and cloud computing

↓

AI models

↓

Applications and AI agents

↓

You

Power can exist at every layer.

A company doesn't necessarily need to own the AI model to wield enormous influence over the ecosystem. Controlling scarce computing resources, distribution channels, or access to millions of customers can be just as important.

That's why asking "Who owns AI?" doesn't quite capture the issue.

A better question is:

Where does control over AI exist?

Open or Closed?

There's another debate unfolding beneath the competition between companies.

How open should artificial intelligence be?

Some leading AI systems are largely controlled by the organizations that develop them. Users interact with the model through an application or service, but they can't simply take the underlying model and modify it however they choose.

There are advantages to this approach.

Centralized control can make it easier to update models, manage security, establish safeguards, and prevent certain kinds of misuse.

But there's another philosophy.

Open and open-weight AI models can give developers, researchers, businesses, and sometimes individual users much greater ability to experiment with and customize the technology.

Meta, for example, has made openness a significant part of its AI strategy and continues to release models and AI tools that developers can use and modify. In August, it released the weights of an agentic model designed to run locally on consumer hardware, illustrating how some AI capabilities can potentially move from giant cloud systems onto devices people control themselves.

Open systems can encourage competition and innovation. They can also make centralized safeguards harder to enforce.

And therein lies one of the fundamental questions surrounding AI:

Should powerful technology be tightly controlled because of what it can do—or widely accessible because of who might otherwise control it?

There isn't an easy answer.

What Does "Control" Really Mean?

When we talk about controlling AI, we're actually talking about several different kinds of power.

Economic control: Who can afford to build the most capable systems?

Infrastructure control: Who owns the chips, cloud platforms, and data centers needed to run them?

Technical control: Who determines how models behave and what capabilities they have?

Distribution control: Which companies have direct access to billions of users through phones, search engines, social networks, browsers, and workplace software?

Information control: How do AI systems determine what information appears in their answers?

Personal control: What choices do users have about their data, privacy, and which AI systems they use?

Seen this way, the future of AI isn't simply a contest to build the smartest model. It's also a contest over the infrastructure surrounding intelligence itself.

What About Government?

Governments are entering this discussion as well. Their challenge is particularly difficult.

Regulate too little, and consumers could face problems involving privacy, discrimination, misinformation, security, or unsafe systems.

Regulate too aggressively, and governments could slow innovation—or inadvertently make the largest technology companies even more powerful because they're the organizations best able to absorb expensive compliance requirements.

Governments also aren't thinking only about consumer protection.

AI has become an issue of international economic competition and national security. Countries increasingly view leadership in advanced computing, semiconductor manufacturing, energy infrastructure, and artificial intelligence as strategically important.

That means governments are simultaneously trying to encourage AI development, regulate its risks, maintain competition, and protect national interests.

Those goals don't always point in the same direction.

Why Should You Care?

It's reasonable to wonder why any of this matters to someone who simply wants AI to help write an email or plan a vacation.

The answer becomes clearer when we consider where AI may be headed.

Imagine AI systems that increasingly help us:

• Find information.

• Choose what to buy.

• Manage our calendars.

• Teach our children.

• Perform our jobs.

• Navigate healthcare.

• Manage financial decisions.

• Create entertainment.

• Communicate with businesses and governments.

If AI becomes an intermediary between us and more of the digital world, then the rules governing those systems become much more consequential.

• What information does your AI assistant have access to?

• What does it remember about you?

• Who can see that information?

• Can you take your data to another AI?

• Can you choose a different model?

• Who decides what the system will or won't answer?

• And what happens if the AI you rely upon changes its rules?

Those aren't merely technical questions. They're questions about individual choice and power.

The Future Doesn't Have to Be Either-Or

It's tempting to reduce the AI debate to two extremes.

On one side: powerful corporations controlling artificial intelligence.

On the other: completely open systems with no meaningful restrictions.

The future is unlikely to be that simple. We may instead develop an AI ecosystem containing many different approaches.

• Large commercial AI systems.

• Open models.

• Smaller specialized models.

• AI running directly on personal devices.

• Government regulation.

• Academic and nonprofit research.

• Competitive startups.

• Perhaps even new forms of decentralized infrastructure.

Each model involves tradeoffs between innovation, safety, accessibility, privacy, competition, and control.

The goal shouldn't necessarily be eliminating power from the system. Building technologies this complex requires institutions capable of organizing enormous resources.

The more important question is whether enough alternatives exist to prevent any single organization—or small group of organizations—from determining the future for everyone else.

The Bigger Picture

Artificial intelligence often appears weightless.

We type words into a screen and intelligence seems to arrive from somewhere in the cloud. But there is nothing weightless about it.

Behind every AI response are chips, electricity, data centers, networks, software, capital, companies, researchers, and increasingly governments.

Understanding that infrastructure changes how we think about AI.

The most important competition may not simply be over who builds the smartest chatbot. It may be over who controls the layers underneath it—and how much choice the rest of us retain as artificial intelligence becomes more deeply embedded in everyday life.

That makes "Who controls AI?" one of the defining technology questions of the coming decade.

But an equally important question comes next:

What happens when these systems stop merely answering our questions and begin acting on our behalf?

That's where we'll go in the final installment of AI for the Rest of Us.

Stay ahead of the curve with the latest in Web3 culture and innovation. Subscribe to Hashed Out for exclusive insights, case studies, and deep dives into the decentralized future.

Help Grow Hashed Out And Get Rewarded With Premium Content & Merchandise

If you believe in a more open, fair internet — help us build it, one reader at a time.

Web3 adoption starts with curiosity. Share Hashed Out with someone who’s ready to explore.

You’re not just sharing a newsletter — you’re inviting someone into the future of digital life.

Refer 3 friends and unlock premium content. The more you share the more rewards you unlock, including Hashed Out mugs or tote bags, and exclusive community memberships.

Other Articles In This Series:

• Hashed Out V78: Part 1: Living With AI

• Living With AI: What It Can Do, What It Can't, and Why It Matters

• Hashed Out V79: Part 2: Can You Trust AI?

• Can You Trust AI?: How to Separate Helpful Answers from Confident Mistakes

• CHECK OUT OUR PREVIOUS EDITIONS