Are safe areas an effective option to protect civilian populations from mass atrocities when they are targeted by their own state? Safe areas disappeared from the international lexicon following the failures in Bosnia and Rwanda. But they are now receiving a second look as a way of responding to mass atrocities without full-scale military intervention. This article argues that the earlier generation of safe areas failed not due to their size or cost, but rather because of problems inherent with their underlying logic. Safe areas were based either on logics of consent or the presence of a credible military force. Hybrid safe areas (such as in Bosnia) were based on neither of these, but instead relied on the legitimacy inherent in the UN Security Council. Crucially, in cases where civilians were being directly targeted by belligerents, both hybrid and consent-based safe areas collapsed. This has direct ramifications for present discussions around the Protection of Civilians agenda and the Responsibility to Protect doctrine.