During the United Nations General Assembly, a coalition of politicians, scientists, and activists introduced an initiative called "Global Call for AI Red Lines". The aim is to establish international prohibitions on the most dangerous applications of artificial intelligence by the end of 2026.
The document has already garnered over 200 signatures, including notable figures such as former Irish President Mary Robinson, ex-Colombian President Juan Manuel Santos, and prominent researchers like Geoffrey Hinton and Yoshua Bengio, often referred to as the "godfathers of AI".
Participants are urging governments to agree on fundamental restrictions that would prevent "universally unacceptable risks". While specific guidelines are not detailed in the document, examples of potential "red lines" are provided, including prohibitions on using AI for nuclear weapons control, mass surveillance of citizens, or creating systems that are impossible for humans to halt.
The organizers propose that the future agreement should be based on three key principles:
- a clear list of prohibited practices;
- independent verification and auditing mechanisms;
- the establishment of an international body to oversee compliance with the agreements.
However, the final determination of boundaries and procedures is left to the discretion of states. The initiators recommend holding special summits and working groups to align positions.
The United States has already committed to "not allowing AI to control nuclear weapons", a pledge made during the Biden administration. However, intelligence officials from the Trump administration have expressed dissatisfaction with companies developing AI being restricted from using their technologies for domestic surveillance. Such issues may complicate the adoption of global security decisions anticipated by AI experts.
