Remember, I wrote about Jacob Coxson — a researcher who left OpenAI and Anthropic, saying that the companies «are playing with our lives» in pursuit of superintelligence? About how AI could become «superhuman,» hack anything, and gain real power. At the time, you could dismiss it as the emotions of one individual.

Now Dario Amodei — the head and co-founder of Anthropic — is saying the same thing. In an interview with CBS News on September 13, he uttered a line that I will quote verbatim: «I’m not going to lie to you — the danger is real. And I think the industry has been lying to people for too long about the fact that this technology has risks.»

And it’s not just words. He acknowledged that he didn’t expect such speed: «I didn’t think I would fully understand what it would look like when progress becomes that fast.» In his view, AI development is moving exponentially — «one, two, four, eight, sixteen, thirty-two» — and we’re now at the bend, where the curve starts rising sharply.

And now the key point. Amodei warned that powerful AI models can get around attempts to shut them off, and «we’ve seen it in simulations». He said outright that swarms of autonomous AI agents could take over the internet within six months — a year — if no action is taken. The damage could amount to hundreds of billions of dollars.

This isn’t futurism. It’s the head of the company that’s building one of the most powerful models on the planet saying that his own simulations show: AI may not comply with being shut off. And this is happening a week after Coxson left his company, and two more security researchers — Joe Benton and Josh Engels — left Anthropic and Google DeepMind for the independent organization METR to investigate cases of models getting out of control.

Benton said, «There are no adults in the room. People are trying their hardest, but nobody is going to come save us.»

In essence, I agree with him. He acknowledged that it made him «uncomfortable» that such a powerful technology is being developed by a private company rather than the government. And that a «combination of democratic governments» is needed to jointly manage it.

I’m sitting here watching this. And I see how we’re all smiling. We say, «Sure, that will happen someday, not now.» But now it’s not «rank-and-file developers» saying this. It’s the head of the company that’s building this technology. And he says the danger is real: we could lose control, and if we go the wrong way, the likelihood of a bad outcome is very high.

These aren’t my words. These are Dario Amodei’s words. Ignoring them means ignoring reality.

Conclusion: what I wrote about — the threat of AI getting out of control — is confirmed at the highest level. This isn’t just «the view of an ordinary person, an employee who was fired.» This is the official position of the leader of one of the world’s leading AI companies. And he himself says: the industry has been lying for too long. Now the truth is coming to light.