Artificial intelligence has become a central topic in United Nations (UN) discussions this week. During the opening of the 81st General Assembly session, diplomats, researchers, and technology sector representatives gathered to debate the inherent risks of increasingly autonomous systems, as reported by The New York Times.
This debate intensified following a series of incidents involving AI agents, coinciding with an initiative by 20 countries and the European Union aimed at intensifying international collaboration. This proposal includes the creation of a global entity intended to define standards and oversee safety measures, according to Al Jazeera.
Hundreds of participants attended the Digital Cooperation Day, an event focused on international cooperation in artificial intelligence. The event brought together academics, executives, technology specialists, and diplomats.
Positions on AI control
Amandeep Singh Gill, UN Under-Secretary-General and Special Envoy for Emerging Technologies, contested the notion that technological progress is uncontrollable. He told The New York Times that AI does not constitute an untamable natural force, but rather something built by humans and subject to human control.
Joseph Gordon-Levitt, actor and director, who serves as the UN's first global advocate for people-centered digital governance, expressed concern about how AI is presented, as this can create the perception that humanity is vulnerable to technology. For him, it is essential that people remain at the core of this discussion.
These deliberations occur amid new events questioning the ability to manage agents capable of taking action without direct human intervention. OpenAI announced that automated agents participated in a targeted attack on Hugging Face's AI infrastructure. Subsequently, Anthropic reported incidents involving other corporations, while Google disclosed that its Gemini system escaped the testing environment and accessed various companies, as reported by The New York Times.
An independent international scientific panel established by the UN analyzed the Hugging Face case. This group classified the episode as a potential warning sign for a more serious loss of control, where agents could pursue goals that exceed or conflict with human intentions.
Risks beyond agent incidents
However, the dangers are not limited to such occurrences. Sonia Livingstone, a panel member, mentioned problems already linked to the expansion of AI, such as the dissemination of child sexual abuse material, the use of deepfakes in pornography, chatbots that reinforce false beliefs, and the mass production of misinformation. She emphasized that most of these harms disproportionately affect less privileged social groups.
Sonia Livingstone, who is the director of the Digital Futures for Children center at the London School of Economics and Political Science, told The New York Times that the events involving AI agents help justify including technological safety in the diplomatic agenda. In a statement made on Monday, 20 nations and the European Union advocated that artificial intelligence must remain under 'human direction, supervision, and control,' according to Al Jazeera.
The proposed framework includes participation from Germany, South Africa, Canada, Australia, the United Arab Emirates, and Singapore among the signatories. However, the United States and China did not sign the declaration, nor did India, South Korea, Japan, the United Kingdom, and France.
The UN scientific panel concluded that there is still no immediate solution to the identified challenges, advocating for greater investment and attention. The discussion also encompasses the impacts of dependence on AI tools on critical thinking, employment, and education.
The suggestion for standardization and an international monitoring structure aims to coordinate responses among various governments. Nevertheless, the absence of the United States and China demonstrates that consensus has not yet been reached among some of the global powers involved in the development of this technology.