Everyone has an opinion on AI right now, and a lot of them are noise. Some say it will annihilate us. Some say it will save us. Both camps get a lot of airtime and neither seems to explain much.
I’ve been thinking about this for a while and I wanted to put some order into it. Not a prediction. Just what I see when I look at who is talking, what they are actually doing, and why the two rarely match.
I see three patterns.

The Oppenheimers

Oppenheimer led the Manhattan Project that built the bombs dropped on Hiroshima and Nagasaki. He saw the damage, got shaken, and told Truman to exercise restraint. In private.
Then the USSR got the bomb. Only then did he argue the super bomb should never be built. Still in private. Still classified.

He went public in 1953. Eight years after Hiroshima. That is the Oppenheimer everyone remembers, and it is the last version of him, not the first.
Eight years of knowing, and saying nothing where it counted. By then his legacy was the thing at stake, not the bomb.
I have huge respect for scientists who build things and push them far. But no one will make me believe that when you are developing something, its dangers are not clear to you. These people are visionaries. That is the whole point of them. A visionary sees where the thing goes.

The pattern I see is the same one today. They become vocal about the dangers of AI only after collecting millions of dollars for making AI come true. At no point before did they judge that the dangers they saw needed to be said out loud. It feels strange to me.

The till-the-end-no-matter-what

People tell me AI companies ignore safety because of profit. I don’t buy it. Most of them are still burning capital rather than making money. OpenAI lost over twelve billion dollars in a single quarter this year. That is not a business cutting corners for margin. That is something else.

I think it’s the space race, the race to the poles, the arms race. It is not about profit but more about who gets there first, never stopping for one second to ask whether it’s relevant, or whether it should be done at all.
Look at how the race is funded. Silicon Valley pours millions into anything that moves, drops it, counts the loss, and writes the next cheque. Losses are not a failure of the model, they are the model. China runs state-directed money on a twenty-year clock, most of it into early stage, against targets set in a five-year plan. Russia isn’t chasing the frontier at all, it’s building something it can control and keep running under sanctions. Different clock, different measure of winning.

The Americans say they won the space race because they put men on the moon, and it was an incredible achievement. But winning a race means beating your competitor. The moon was never the Soviets’ main business. They were racing a different race, and they were far more interested in what you could do from orbit.
That is the danger of the till-the-end-no-matter-what. The blindness to anything that isn’t the race. Believing it’s life or death to win, never stopping to see that no one is racing them, and that most of the time they are running full speed at a concrete wall with no brakes.

The synthetists

For them AI is a friend. They want it in every home, always listening, always watching, always ready with advice and counsel.

Some have had virtual girlfriends and boyfriends. Some simply believe everyone should. They fund humanoid robots, push brain chips, always-on devices, more and more human-machine contact, blurring the border between man and machine as far as it will go. Every so often one of them goes public to say the model is conscious, or that AI will understand you better than people do, that AI is your friend.

This is Pygmalion and it is hubris. Pygmalion fell in love with the creature he carved. Hubris is forgetting there is a limit, and that crossing it has always had a price. One is loving your creation. The other is overreach. These people are doing both at once, and overreach has never once ended well.

Humans have a long history of trying to create something because they never got over the fact that they didn’t make themselves. So they chase the dream, obsess over it, and never once stop to ask whether the thing they are chasing is a very bad thing to catch.

Where I land

Regulating AI is legitimate. The people doing the loudest regulating are not always the ones with the credibility to do it.

Every technology humans ever invented came with trade-offs, and the trade-offs are what triggered the rules. The industrial revolution brought cheaper goods and more purchasing power, and it brought pollution and waste. The digital revolution put intelligence in a box and expanded what we could create and calculate, and it also gave us social media, online harassment and propaganda at scale. Railway safety, workers’ rights, checklists in cockpits. All of it written after the damage.

AI is one of the few times we did it the other way round. Europe passed its AI Act before the catastrophe. China made labelling of AI-generated content mandatory before anyone was forced to. For once the rules arrived ahead of the tragedy. And then the deadlines moved. The high-risk rules that were meant to apply in 2026 now apply in 2027 and 2028. Too burdensome, not ready, bad for competitiveness.

Here is what I think we should do instead. Treat AI as a strategic and sovereign matter, the way we treat energy, transport and defence. Build and deploy the systems we have judged safe, on our own timetable, and keep the guardrails and the common sense that come with anything strategic. Not a world where AI has the last word. A world where AI stays the tool it is and should remain. That is the line for me. AI decides nothing on its own. A human validates and signs.

So yes, raising concerns about AI moving too fast is legitimate. What isn’t legitimate is excusing your own recklessness by pointing at a competitor who was never racing you.