More than a dozen of the world's leading AI firms have made fresh safety commitments at a global summit in South Korea.
"In the most extreme circumstances, the companies have also committed to 'not develop or deploy a model or system at all' if mitigations cannot keep risks below the thresholds," the statement added.The definition of these thresholds will be decided ahead of the next AI summit, due to be hosted by France in 2025.
In addition to safety, the Seoul summit will discuss how governments can help spur innovation, including into AI research at universities.