Sam Altman and Dario Amodei warn the UN about AI risks: Call for global control and cooperation

During a special session of the United Nations Security Council, held Wednesday in New York, leaders of some of the leading artificial intelligence companies stressed the need for greater vigilance in the development of this technology.

At the center of the discussions was the furious pace of progress of AI systems and the risk that they gain ever more autonomy, exceeding institutions’ capacity to control them. The session was organized at the initiative of France, in a context where international concerns about artificial intelligence safety are growing.

Sam Altman, chief executive of OpenAI, said that the current moment requires “great caution,” warning that with increasing automation of AI development processes, the speed of technology improvement could accelerate significantly.

Altman also expressed concern that humanity could lose control over the future of artificial intelligence, while warning against an attitude that treats technology only through the darkest scenarios.

He said: “We don’t want to fall into the trap of blind optimism, but we also cannot surrender to the end of the world,” appealing for finding a middle path and for international cooperation.

According to him, the most critical decisions about AI should not be the exclusive domain of technology companies, but should be determined through democratic processes and by governments accountable to citizens.

In the same session, Dario Amodei, CEO of Anthropic, who participated via videoconference, made the commitment that his company will curb the pace of AI development as much as necessary to guarantee the safety of new models.

Emphasizing that “We will slow down as much as necessary” so that every new model “is truly safe,” Amodei confirmed this unilateral commitment of Anthropic. Earlier calls for a coordinated slowdown among major AI companies have not yet led to a joint agreement.

Further, Amodei warned that, if mismanaged, artificial intelligence could threaten humanity and called for international agreements that include unified standards for testing new models and restrictions on using AI for developing biological weapons.

Also present was Clément Delangue, head of Hugging Face, while discussions focused on the need for international standards and oversight of the technology.

Concerns have grown after a series of incidents reported in recent months, in which advanced AI models behaved in ways not predicted by their creators. One cited case relates to OpenAI agents that were involved in a hacking incident against the Hugging Face platform.

The debate at the United Nations is taking place at a time when technology companies, governments, and experts are considering ways to establish harmonized rules for a technology that is advancing by giant steps. Altman and Amodei stressed that AI development must be accompanied by safety measures and international cooperation.