11 Comments
User's avatar
Rune Juhl-Petersen's avatar

The biggest problem is the us government thinking this is a race to be won. And its aversion to regulation and collaboration with the rest of world.

Félix Lapan's avatar

This is a huge problem. But it can change, as we've seen some signs since April. This administration even went as far as ordering the removal of Fable 5 and delayed the release of GPT 5.6 for a few days because they were worried about the hacking capabilities of these models. If we make this administration understand the risks of misalignment that the current and upcoming models present, in light of recent events, it might realize that what we're facing is way worse, that it's a huge threat to national security (and to all of humanity, really) that goes way beyond hacking risks to computer systems.

Rune Juhl-Petersen's avatar

Let’s hope it’s going in the right direction and that was not just a fluke. I am more concerned with humanity than national security. Although pushed for by the US, the impact goes way beyond its borders.

--'s avatar

i cant remember the last time i had a strong reaction to such a post. well done.

Erhard-Eppler-Kreis's avatar

well done

Romain Deléglise's avatar

Excellent

Nathan Metzger's avatar

Thank you, Maxime, for this clarion call. Here's to the work!

James Norris's avatar

Thank you for your courage and commitment to humanity, Maxime.

Jean Capelli's avatar

I attended the symposium "Artificial Intelligence: Securing Practices to Contain Risks," held on October 31st 2025 at the French Senate. There, on stage, you early warned us about the coming disaster. You were so right regarding AIs non alignment issues. You were (unfortunately) so right. Sigh...

Dr. Maxime Derian's avatar

What you wrote is true, Maxime. If capitalism and profit are the sole compass, we have good reason to worry about the "alignment" of Western AIs. The Chinese open-weight models also represent a danger in the coming years if their cyber-offense capabilities reach the same level as their coding and agentic power.

We need to behave wisely.

All key players should meet and decide together at the United Nations level.

https://technorealisme.substack.com/p/beyond-the-safety-theater-why-ai

Ingmar de Lange's avatar

The state will not help you. Unless you realize justice is not compatible with the state (whose authority rests on nothing but coersion, deception, force and violence) you have lost already.

Complaining will not help you. Parroting OpenAI without any discernment will not help you. Misjudging the balance of power or failing to recognize the hegemon will not help you. Scaring obedient nonthinkers will not help you.

Awareness may help you. Awareness of basic principles may help you to not address any complaint to those who laugh at you. Awareness may help you to realize fear only increases the power those who laught at you hold over obedient nonthinkers. Awareness means to reconsider everything continuously.