At the UN General Assembly, a group of politicians, scientists, and public figures presented an initiative called the "Global Call for AI Red Lines." Its goal is to develop international bans on the most dangerous scenarios for the use of artificial intelligence by the end of 2026. This is reported by Gizmodo.
The document has already been signed by more than 200 people, including former Irish President Mary Robinson, former Colombian President Juan Manuel Santos, and Jeffrey Hinton and Joshua Bengio, who are called the "godfathers of artificial intelligence." The signatories also include prominent scientists and public figures.
The participants call on governments to agree on basic limits that would prevent “universally unacceptable risks.” There are no specific rules in the document, but examples of possible “red lines” are given. These include a ban on the use of AI to launch nuclear weapons, use in mass surveillance of citizens, or the creation of systems that cannot be stopped by humans.
The organizers propose that the future agreement be based on three principles:
- a clear list of prohibited practices;
- independent verification and audit mechanisms;
- creation of an international body to monitor the implementation of agreements.
At the same time, the final definition of boundaries and procedures is left to the discretion of the states. The initiators recommend holding special summits and working groups to coordinate positions.
The US already has a commitment to “not allow AI to control nuclear weapons,” dating back to the Biden administration. However, intelligence officials in the Trump administration have reportedly expressed frustration that AI companies are not allowing their work to be used for domestic surveillance. Such cases could complicate the global security decisions that AI experts are calling for.