“AI’s Prisoner Dilemma”

Everyone can agree that safety standards and guardrails would be better for humanity. But everyone also has an enormous incentive to cheat.

Source