Black Hat Europe 2024: Can AI systems be socially engineered?

Could attackers use seemingly innocuous prompts to manipulate an AI system and even make it their unwitting ally?