Will AI end humanity?
Fear of superintelligent AI may be legitimate, but it is also useful to those who need to fund an increasingly expensive race. The immediate risk may lie less in the models than in the economy built around them.
Jacob Coxon, an AI researcher, resigned from Anthropic this week, saying that AI companies are in a runaway race towards superintelligence and are therefore gambling with our lives.
I have written about this here before. And I stand by my view: we need an international agreement on AI.
The idea that a superintelligent machine could be humanity’s last invention, leading to extinction, is not new. I. J. Good was already talking about it in 1965. Stephen Hawking warned about it in 2014. Geoffrey Hinton, the "godfather of AI" and a Nobel laureate, has talked about little else for years.
What is new, and strange, is that Elon Musk and Sam Altman have both publicly agreed with Dario Amodei (Anthropic) when he says we need to slow down.
Even stranger:
- Coxon has only had an X account for a few months
- he presented no new data to support what he said
- people who leave these companies generally do not speak this freely; contractually, it can be very costly
All of this comes in the year SpaceX went public, Anthropic is preparing to be next, and OpenAI postponed the IPO it had planned for this year.
The truth may not make headlines as good as "the end of humanity". AI is advancing quickly, but mass adoption, especially in companies, has not kept up with what business models predicted. Workflows need to be rethought, people need training, trust needs to be built and internal resistance needs to be managed. All of this takes a long time.
Then there is China, which is releasing models 60 to 90% cheaper. Good enough for everyday use. And it is releasing open-weight models, which can be used at no cost beyond energy and hardware. OpenRouter says that in 2025, 70% of the models used on its platform were American. In 2026, 30%.
Add to that the elections in the US and the fact that much of the electorate opposes the construction of data centres.
So how does all this buzz about the end of humanity help?
War. Oil and inflation rise, so interest rates rise. Big Tech’s capex for 2026 is $754 billion, not counting what has already been invested and the interest paid on that money.
These recent statements give the impression that Big Tech companies do, after all, care.
Superintelligence — AGI — has been the carrot for investors in these companies. Every year, Altman says it is "almost here". And the story is no longer landing in the same way. At the launch of the latest model, Altman himself and Jensen Huang (Nvidia) even said AGI had arrived.
Spoiler alert: it has not.
If it is a dangerous technology, perhaps it is important for the State to become a partner and put fresh money into companies that need it so badly.
If it is dangerous, perhaps it makes sense to regulate its use and ban access to Chinese models, leaving American companies as the only ones authorised in the market.
From what I have followed in AI, the characters in *Succession* are ethically beyond reproach compared with the real-life characters in this new world.
So, stay calm. Right now, if AI ends us, it will be through the economy.