The AI Safety Paradox: When 'Safe' AI Makes Systems More Dangerous

AI safety is becoming an established field and science. Yet by focusing only on the safety of individual models, AI labs may actually be making the conditions in which they’re deployed less safe.