AI Wars Summary: Who Really Owns Your AI?

Your team uses an AI tool every day. It writes drafts, answers customers and helps with code. It feels like something your company owns, like a laptop or a software licence. But what if the company that runs it changes the price next month, or limits who can use it? Most leaders have never asked what they would do.

This episode, AI Wars, argues that you probably do not own your AI. You rent it. And the rules of that rental are being shaped by governments and a few large firms.

What is AI Wars about?

AI Wars: Why You Don’t Own Your AI is an original brief by Amit Jaju. It is not a published book. It is a researched talk for business leaders, and it was made for a viewer in India who runs or advises a company.

The brief does not treat the AI race as a contest between chatbots. It treats it as a contest between countries. The questions are who has the chips, who has the energy, who has the money, and who has the right to operate. The brief says that whoever controls these four things decides who can use the most advanced AI, and on what terms.

This matters now because many firms have built products and daily work on top of AI tools from a small group of suppliers. The risk is easy to miss because everything works well, until the day it does not.

Why did the United States lead?

The first section describes the scale of American spending. The brief says the large US technology firms were spending around 725 billion dollars in a single year on AI infrastructure. It also mentions a planned 500 billion dollar data-centre project called Stargate, and Nvidia, the chip maker, becoming the first company worth 5 trillion dollars.

Money, machines and skilled people stacked together made a wall that looked very hard to climb. For some time, the best AI in the world was American, and everyone else simply rented it.

Picture a manager in Mumbai. The tools her team uses every morning run on systems she does not control and cannot inspect. That seemed fine as long as the supply was steady.

How did China change the picture?

The brief says a Chinese lab called DeepSeek released a model close in quality to the best American ones. It gave the model away free to download. According to the brief, the news wiped about 588 billion dollars from Nvidia’s value in a single day. That was the largest one-day loss for a company in stock-market history.

The brief calls China’s approach smart. Instead of selling AI like premium fuel, it gives models away, so that many developers build on them and use Chinese models as a standard.

The brief also mentions a claim from American labs. They say some Chinese labs trained their models by copying outputs from American ones, a process called distillation. It stresses that China has not admitted this, no court has ruled, and the claim is unproven.

Does “open” mean you can use it freely?

This is the most practical lesson for leaders. “Open weight” means you can download a model. It does not mean you can do what you like with it.

The brief gives examples. Meta’s Llama needs a special licence from Meta past a certain size. Google’s Gemma models, which are free to download, come with terms that let Google restrict use. Meanwhile, several Chinese models come with simple, permissive licences.

There are legal points too. Under the European Union’s AI law, a firm that heavily fine-tunes a model can become its legal “provider” and take on duties. The US Copyright Office has also signalled that content made only by AI may not be protected by copyright.

So before you build a product on a model, read the licence. Ask what you may do, what you may not, and who owns what it produces.

Can access be switched off?

The brief says yes. It points to the chip war. America limited the sale of its best chips to China. Nvidia took a large loss and later agreed to give the US government a share of some China sales. China then blocked an approved chip at its own border, to push firms to use local chips.

It also says that when OpenAI released one of its newest models this year, it limited access at the request of the US government. The lesson for a business is simple: supply can change because of politics, not only because of price.

Where does India stand?

The brief is frank about India. It says India is one of the two largest AI markets by usage, with about 330 million people a month on ChatGPT alone. It also has the world’s second-largest pool of AI talent.

But, the brief says, India owns no model at the global frontier. The government’s AI mission is funded at about 10,000 crore rupees, but only around 4 per cent had been spent in its first two years. Local models such as those from Sarvam and Krutrim are built for Indian languages on smaller budgets. The brief calls them promising, but not yet at the frontier. It also says barely one in 25 Indian organisations has a serious AI-risk framework.

Europe, meanwhile, writes rules that can reach any firm serving the European market, with fines up to 7 per cent of global revenue. The Gulf states are buying chips and compute with oil money.

How to use this at work and at home

1. List every AI tool your team uses, and who supplies it.
2. Ask if you have a second source. If your main supplier vanished, what is the fallback, and has it ever been tested?
3. Ask your lawyer to read the licence of every open model used in production. Check the limits and who owns the output.
4. Find out where your data goes, and whether you could run a self-hosted model if you had to.
5. Hold one board or team meeting on these three questions before your next AI purchase.

Our take

The strongest part of this brief is its audit. The three questions are simple, practical and free to ask. Few teams can answer them today, so even asking is useful. The point about licences is also valuable, since many people treat “open” as “free”.

There are limits. The brief is made by one person and argues a point of view. Some claims rest on news reports and are time-sensitive, such as spending figures and access limits. These can change quickly, so check current facts before you act on them. The language of “weapons” and “switches” is dramatic. In daily business, the risk is more often a price rise or a changed licence than a sudden shutdown. The brief also says little about the benefits of renting AI, such as low cost and fast setup, which suit many small firms very well.

Still, as a prompt for leaders to look at their AI dependence with open eyes, it does its job.

The animated summary walks through the full argument and the three-question audit. It takes about 12 minutes.