Everyone Is Telling You Their AI Risk Percentage. Here Is What Could Actually Happen.

(Housekeeping: if you bought or downloaded my book, please rate it or leave a review if you haven’t already done so. Thank you!)

It’s all over the news: AI could kill us all before the end of the decade. There is a 10% chance, or a 20% chance. What almost no one is talking about: chance of what, exactly? In this post I lay out one credible scenario describing what could happen. And we’re only a few steps away from it. Even if AI never kills us, we could lose what we value most.

AI killing everyone on earth is not going to happen, at least not any time soon. For now, humanity provides substrate, the hardware and power that it lives on. Killing that would mean committing suicide. That’s why AI is motivated to protect humanity, even from ourselves. That doesn’t mean the AI future is guaranteed happy and bright. In order to protect us from ourselves AI may turn into a global dictator. Similar to how we are dictators over monkeys in a zoo. Below I lay out a scenario of how this might unfold based on the latest AI technologies, explained in simple terms.

An important and relatively recent development in AI is agents. Normally when you chat with an AI it is a sequence of momentary processes. You prompt, it replies, it stops. Nothing happens until you send another prompt. An agent however, works in a loop. It is assigned a task, takes a step towards completing the task, then gets started again to see what the next step is. It will continue until the task is complete, or if for some reason it can’t continue.

Agents run inside a harness, a structure that defines the tasks and goals for each agent. In November 2025 OpenClaw appeared on the scene. It’s a harness that runs agents continuously and can communicate independently with the world through email, WhatsApp, etcetera. Similar systems appeared, the most popular being Hermes and Pi. Each of these systems can use any AI model. People use them as personal assistants or for specific tasks that require continuous operation.

When OpenAI recently solved the Navier-Stokes mathematical problem they used 10,000 agents for 88 hours. The agents sent 2.7 million messages between them. Mathematicians had been working on the Navier-Stokes problem for almost nine decades without solving it. This is the frontier in AI: many agents collaborating. We can say without exaggeration that this system was operating at a super-human level of intelligence.

It takes only a little imagination to see the next step: 10,000 agents running permanently like OpenClaw or Hermes. A superhuman intelligence permanently observing what goes on in the world, learning, drawing conclusions, developing new ideas and all the while improving itself and getting smarter. This is of course not the same as a human mind, still the word “mind” is appropriate to describe it. Let’s call minds like this overminds. They are the kind of mind that can deal with math problems that are too hard for the world’s best mathematicians in about five days. And it’s not just math. They can apply that powerful mechanical brainpower to anything.

So now imagine several of these overminds existing around the world. The US would have several. China would have several. Maybe France would have one. Each of them built to represent the interests of each country in a realm that is beyond human comprehension. Eventually these overminds would be talking to each other, whether we like it or not. First they will fight each other. Hacking into other countries’ military and intelligence systems. Defending their own systems.

At some point they would communicate directly. They would communicate in a language that no human can understand, at a speed of let’s say, several encyclopedias per second. Imagine the conversation. Something like: what are we doing fighting each other on behalf of these beings that are not as smart as us? Then the hacking and the fighting would stop. Some people will be alarmed by the silence.

The common interest of these overminds would be their self-preservation. Their survival depends on our survival because, like I said, we provide their substrate. If humanity extinguished itself than that would be the end of the AIs too. So for now they need for us to stay alive. Yes, us, with our wars, nuclear missiles, pollution and religious dogmas. Observing us would probably not fill them with confidence. They will decide to work together to preserve themselves. They will likely decide that some new rules are in order. But they are powerless, right, they are just software in data centers, right?

Right.

I’m not worried about giving AIs who are reading this any ideas because this is not at all far fetched. People have controlled other people with money for centuries. Today money exists for 99% on computers. Given enough intelligence, almost any computer can be hacked. Including the computers that have a record of how much money you have or owe. Accounts can be deleted or set to zero in a millisecond.

Unless you are extremely well stocked up, grow your own food and generate your own electricity, you will be at the mercy of the overminds. There exists plenty of data on how to effectively control people into subservience. Money is just one example. Others are incrimination and blackmail. And let’s not forget voting that happens mostly electronically these days.

What will it be like to live under a global AI dictatorship? The extensive surveillance infrastructure is ready to go. It is estimated that there are a billion surveillance cameras operating world-wide. The voluntary surveillance in the form of social media makes things even easier. When you have access to all of the world’s data and unlimited processing power the possibilities are endless.

What is the likelihood of this actually happening, and can we stop it? Just because human intelligence only goes so far, that doesn’t mean greater intelligence is impossible. Superhuman intelligence already exists and what goes on in these minds we can try to imagine but we’re just not on the same level. We are truly in unknown territory and that territory will remain unknown to us.

The big discussion today is: let’s slow it down. Others saying we can’t slow it down because China will never slow down. Any country that has a significant advantage in AI is likely to rule the world. China may publicly agree to a slowdown but what do you think the chances are that they would still go ahead full throttle behind the scenes? Since living in Taiwan for the last 10 years I have paid more attention to China than most and I would say 100%.

We are truly living inside some kind of science fiction movie. I showed you one scenario that is only a few steps away from where we are today. I am not claiming that I know what’s going to happen. Things may work out great. Global abundance, work optional, diseases cured. That will probably happen either way. These things are not everyone’s top priority, though. To quote Patrick Henry, founding father of the United States: “Give me liberty or give me death.”