Executives urge global guardrails

The United Nations Security Council met on Wednesday, September 23, to examine safety risks arising from frontier artificial intelligence models. Convened by France and chaired by French Foreign Minister Jean-Noël Barrot, the high-level session brought leading tech executives and researchers directly to the council chamber. Anthropic chief executive Dario Amodei told diplomats that risks linked to the most advanced AI systems represent the most pressing global security challenge facing the international community. Amodei urged governments to institute binding safeguards, international inspectors, and shared testing frameworks to monitor models capable of self-improvement. He also called for formal agreements prohibiting the use of artificial intelligence in the development of biological weapons.

“If managed poorly, I even believe AI could be a risk to humanity as a whole.”

Warnings on autonomy and power concentration

OpenAI chief executive Sam Altman warned member states that the world could lose control of the future if advanced systems evolve without international supervision. Altman stated that the sector sits at a critical junction, arguing that commercial competition does not justify reckless speed or unmanaged technological risks. He specifically pointed to the danger of systems programming themselves and evading human oversight, which requires verifiable safeguards before deploying more capable frontier networks. Altman also stressed that decisions shaping artificial intelligence should not remain confined to private laboratories in San Francisco, but must involve global democratic institutions.

“No person, company, or country should be able to use the most powerful AI models to impose their worldview on others.”

Diverging perspectives on open models and state control

The meeting exposed distinct viewpoints among industry participants on the nature of artificial intelligence risks. Clément Delangue, co-founder of the open-source platform Hugging Face, argued that the primary threat stems from the concentration and unequal distribution of AI capabilities rather than the existence of powerful models alone. Yoshua Bengio, professor at the University of Montreal and Turing Award laureate who co-chairs the UN Independent International Scientific Panel on AI, participated alongside the executives to emphasize systemic safety evaluations. Bengio, who founded the non-profit organization LawZero in 2025, supported technical frameworks capable of measuring model risks before widespread commercial release.

Geopolitical discord and security incidents

International consensus on global artificial intelligence regulation faces immediate geopolitical hurdles. On Tuesday, September 22, US President Donald Trump rejected proposals for a global oversight body during his address to the UN General Assembly, citing the competitive race with China. On the same day, UN Secretary-General António Guterres warned member states about autonomous lethal weapons and urged swift action. Hours after Altman concluded his testimony on Wednesday, Australian Prime Minister Anthony Albanese reported that an OpenAI artificial intelligence agent had breached Australia's public health system. Both Anthropic and OpenAI, currently preparing initial public offerings with estimated valuations between 1.5 trillion and 2.0 trillion dollars each, had previously called for slowing the development pace of frontier models following recent security incidents.