Ahead of the second AI Security Summit today and tomorrow, May 21 and 22, in Seoul, leading scientists in the field are calling on world leaders to take stronger action on the dangers of artificial intelligence, finding that progress is insufficient since the first Summit held six months ago.

In an article published in the journal Science, 25 scientists from the US, EU, UK and China report that not enough has actually been done to protect us from the dangers of technology. As they point out, it is imperative that world leaders take seriously the possibility that within the current decade or the next, extremely powerful general artificial intelligence systems will be developed, surpassing human capabilities in many critical areas.

At the same time, although governments worldwide have made some efforts to introduce initial guidelines, their moves do not correspond to the possibility of rapid progress that many experts expect.

Furthermore, they find that current AI security research is lacking, with only 1-3% of AI publications addressing security. They note that neither the mechanisms nor the institutions exist to prevent the misuse of technology.

Scientists are calling on governments to set up fast-track expert bodies to oversee artificial intelligence, with far more funding than expected. Also, mandate much more rigorous risk assessments, but also require AI companies to prioritize security and demonstrate that their systems cannot cause harm. Finally, to take a leading role in creating a regulatory framework.

The article is co-authored by, among others, historian and philosopher Yuval Noah Harari, the world’s most widely cited computer scientist and AI pioneer Jeffrey Hinton, and most widely cited AI security researcher Dawn Song.