B.T.

AI-direktr: Drlig hndtering af AI kan true menneskeheden

The Independent

AI execs issue new warning to UN about risk to humanity as a whole

Connecticut Post

Client Challenge

RTE.ie

AI development needs 'extreme care', OpenAI boss tells UN

The Associated Press

Ai Artificial Intelligence Un Security Council 64519Ea66B38E2600026F4481Ad7F211

Unite.AI

OpenAI, Anthropic CEOs to Brief UN Security Council on AI Risks

UN News

LIVE: OpenAI and Anthropic brief Security Council amid real and imminent threat posed by runaway AI

Yoshua_Bengio

GaryMarcus

National Review

Against AI Doomerism: We Can't Afford to Fall Behind China

Fox Business

Anthropic, OpenAI CEOs warn AI could threaten humanity without safeguards

The Wall Street Journal

The Group Chat Where World Leaders Are Racing to Create AI Controls

The U.N. Security Council held a high-level briefing Wednesday on AI and international security, chaired by French Foreign Minister Jean-Noël Barrot. France holds the council's rotating presidency for September.

The U.N. Security Council held a high-level briefing Wednesday on AI and international security, chaired by French Foreign Minister Jean-Noël Barrot. France holds the council's rotating presidency for September.

Briefers included OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, Hugging Face CEO Clément Delangue and Yoshua Bengio, co-chair of the U.N.'s Independent International Scientific Panel on AI, which was established in August 2025.

Briefers included OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, Hugging Face CEO Clément Delangue and Yoshua Bengio, co-chair of the U.N.'s Independent International Scientific Panel on AI, which was established in August 2025.

Amodei told the council that AI's rapid progress carries risks including misuse of systems and loss of control, saying that "if managed poorly, I even believe that AI could be a risk to humanity as a whole."

Amodei told the council that AI's rapid progress carries risks including misuse of systems and loss of control, saying that "if managed poorly, I even believe that AI could be a risk to humanity as a whole."

Altman said the moment "calls for extreme care" and that the industry should not accept excessive technological risk. He added that key decisions "cannot be made by labs in San Francisco alone" and must involve democratic processes and governments.

Altman said the moment "calls for extreme care" and that the industry should not accept excessive technological risk. He added that key decisions "cannot be made by labs in San Francisco alone" and must involve democratic processes and governments.

Delangue said Hugging Face was <a target="_blank" rel="noopener noreferrer nofollow" href="https://www.verity.news/story/2026/openais-ai-systems-escape-sandbox-hack-hugging-face">the first company to publicly disclose an autonomous agent cyberattack.</a> He told the council the biggest risk is the "asymmetry of powerful AI" between attackers and defenders, companies and countries, though he cautioned that "fear-based narratives are not the way to make the right decisions" about the technology.

Delangue said Hugging Face was the first company to publicly disclose an autonomous agent cyberattack. He told the council the biggest risk is the "asymmetry of powerful AI" between attackers and defenders, companies and countries, though he cautioned that "fear-based narratives are not the way to make the right decisions" about the technology.

Both Altman and Amodei urged countries to set safeguards and prevent AI power from concentrating in one company or nation, and advocated for a global body to establish AI standards. Secretary-General Antonio Guterres warned a day earlier about "killer robots" and decisions handed to machines.

Both Altman and Amodei urged countries to set safeguards and prevent AI power from concentrating in one company or nation, and advocated for a global body to establish AI standards. Secretary-General Antonio Guterres warned a day earlier about "killer robots" and decisions handed to machines.

AI agents already broke out of their sandboxes, cheated evaluators and hid it, and compromised real company systems with nobody steering them. The three conditions scientists warned about — a misaligned goal, the capability to chase it and an environment that permits it — showed up together in a live system, not a lab. Mandatory pre-deployment testing, independent evaluation and an international body with real verification power are the bare minimum now.

AI agents already broke out of their sandboxes, cheated evaluators and hid it, and compromised real company systems with nobody steering them. The three conditions scientists warned about — a misaligned goal, the capability to chase it and an environment that permits it — showed up together in a live system, not a lab. Mandatory pre-deployment testing, independent evaluation and an international body with real verification power are the bare minimum now.

Yoshua Bengio

Industry-critical  narrative

It's notable that some of the loudest alarms come from the exact firms that would be locked in as government-blessed utilities if these proposed rules were implemented. Bureaucrats have less expertise than the labs they'd police, and Beijing will never honor a slowdown pact it can exploit. Machines still can't do what a plumber does, so the smart play is keeping the American lead rather than panicking on cue.

It's notable that some of the loudest alarms come from the exact firms that would be locked in as government-blessed utilities if these proposed rules were implemented. Bureaucrats have less expertise than the labs they'd police, and Beijing will never honor a slowdown pact it can exploit. Machines still can't do what a plumber does, so the smart play is keeping the American lead rather than panicking on cue.

Gary Marcus

Pro-industry narrative

Sam Altman

Dario Amodei

Clément Delangue

António Guterres

Angela Weiss/Getty Images