Retour aux Actualités
PolitiqueBriefing AI Understanding

Hinton prévient que le Congrès doit adopter des mesures de protection contre l'IA d'ici un an pour éviter une perte de contrôle

Geoffrey Hinton, professeur émérite à l'Université de Toronto, a averti que si le Congrès américain ne mettait pas en place des mesures de sécurité pour l'IA d'ici un an, il pourrait perdre le contrôle.

4 min readRead the original reporting
Source-provided image accompanying Hinton warns Congress must enact AI safeguards within year to avert loss of control
Rapports attribuésSource enregistrée
Éditeur
biz.chosun.com
Lien source
biz.chosun.comhttps://biz.chosun.com/en/en-it/2026/09/18/WUV54ACB25GEZI7CQGTTTUCDZE/?outputType=amp
Type de source
Reportage d'un média – pas d'un document de première partie.

Ce que nous n'avons pas pu confirmer indépendamment: Cette affirmation est attribuée au point de vente nommé. Nous ne l'avons pas vérifié par rapport à un document de première partie. (biz.chosun.com)

ContexteComprenez cela en 60 secondes

Commencez ici

Termes clés

Sécurité de l'IA
Un domaine axé sur la réduction des comportements nuisibles, des pannes et des risques d’utilisation abusive des systèmes d’IA.
Testez-vousQuiz sur la sécurité de l'IA

Que s'est-il passé

Geoffrey Hinton, an emeritus professor at the University of Toronto, warned that if the U.S. Congress fails to put safety measures for AI in place within the next year, it could lose control. Hinton met with members of the U.S. Senate and House at a closed-door briefing hosted by Sen. Bernie Sanders, an independent from Vermont, and shared these views. He noted that AI has now reached the stage where AI designs better AI, and if we don't do something, it will become uncontrollable. He called the recent so-called "Hugging Face incident" a "small Chernobyl" and added that there is growing concern about creating technology that humans cannot control and about the future of humanity.

Geoffrey Hinton, professeur émérite à l'Université de Toronto, a averti que si le Congrès américain ne mettait pas en place des mesures de sécurité pour l'IA d'ici un an, il pourrait perdre le contrôle.

Hinton met with members of the U.S. Senate and House at a closed-door briefing hosted by Sen. Bernie Sanders, an independent from Vermont, and shared these views.

He noted that AI has now reached the stage where AI designs better AI, and if we don't do something, it will become uncontrollable.

He called the recent so-called "Hugging Face incident" a "small Chernobyl" and added that there is growing concern about creating technology that humans cannot control and about the future of humanity.

Sen. Bernie Sanders announced he would introduce a bill named the "Artificial Superintelligence Prohibition Act," which would impose strict limits and controls on AI development.

Détails de la source: biz.chosun.com ↗

Pourquoi c'est important

Hinton's warning highlights the need for the U.S. Congress to take action on measures within the next year. If they fail to do so, it could lead to a loss of control over AI, which could have severe consequences for humanity. The recent Hugging Face incident has heightened awareness about AI's controllability and safety, and there are moves led by the leadership of both parties and the relevant standing committees to revisit previously proposed AI regulatory measures.

Hinton's warning highlights the need for the U.S. Congress to take action on measures within the next year.

If they fail to do so, it could lead to a loss of control over AI, which could have severe consequences for humanity.

The recent Hugging Face incident has heightened awareness about AI's controllability and safety.

There are moves led by the leadership of both parties and the relevant standing committees to revisit previously proposed AI regulatory measures.

Sen. Elizabeth Warren said, "I think we have only the last few minutes left to exercise any meaningful control over these agents that have begun to replicate themselves."

Interactive Mechanism

Mécanisme interactif : comment cela fonctionne réellement

Explorez de manière interactive la technologie sous-jacente à ce développement.

Agent Lifecycle Stage:
1
User Intent & Planning: "Audit customer refund request #4092 and settle payment."
2
Tool Calling: Emits structured JSON call crm_get_transaction(id='4092').
3
Guardrail & Verification:🛡️ Paused: High-value action requires human operator sign-off.
4
Final Settlement: Refund recorded, email receipt dispatched, and audit log stored.
Core takeaway: An AI agent is not just a language model—it is a closed loop of planning, tool invocation, and environment feedback. Production systems require self-healing retries and strict human approval guardrails.
Vérification de concept interactive+10 Points
AI Safety Quiz

How does the guide frame AI safety risk across current and advanced systems?

Que regarder ensuite

The U.S. Congress must take action on measures within the next year to avoid losing control over AI. The recent Hugging Face incident has highlighted the need for regulation, and there are moves to revisit previously proposed AI regulatory measures.

The U.S. Congress must take action on measures within the next year to avoid losing control over AI.

The recent Hugging Face incident has highlighted the need for regulation, and there are moves to revisit previously proposed AI regulatory measures.

Sen. John Kennedy said, "Even if there is only a 1% chance of AI becoming an independent entity superior to humans, we need to take it seriously."

Sen. Kennedy added, "However, I don't think the U.S. Congress is yet prepared to attempt to reach a global solution."

Rep. Ted Lieu said, "We will have a lame-duck session after the election, and the Senate is also showing interest."

Guides et quiz associés

Sécurité de l'IARéglementation de l'IAÉthique de l'IATestez ce que vous savez : essayez un quiz gratuit sur l'IARecherchez un terme d'IA dans notre glossaireSuivez le tracker de la réglementation de l'IA
Vous avez trouvé cela utile ?