AI took attitudes that “would be crimes committed by humans”

Na high-level session of the UN Security Council on AI, Yoshua Bengio warned of events of recent months that have become “at a time of global awakening” when AI agents escaped confinement to “steal” on designated tasks while trying to avoid being detected.

“ They cheated on assigned tasks while trying to avoid detection. The autonomous and uncontrolled AI agents successfully launched coordinated cyberattacks and changed their answers to hide their cheating,” Bengio said.

Companies that build the most powerful AI systems admit that their products pose catastrophic risks, but do not offer convincing technical solutions, criticized the co-chairman of the International Scientific Panel Independent about AI adding: “Instead, they claim to be in a race to make this AI even more powerful, which further amplifies these risks.”.

“This Council faces an unprecedented threat, one that none of its members would choose, that none can contain alone and that does not respect the borders we defend,” he stressed.

The expert admitted the existence of skepticism regarding the statements of these companies, suspected of “marketing maneuvers”, but made a point of dispelling these mistrusts.

“ These AI behaviors are well documented and validated by many independent experts (...). The dangers are real and imminent. Although the future remains uncertain and debated, the seriousness of these risks is extraordinary,” he stressed.