In response to ethical problems of AI, including its potential to disempower humanity, two normative approaches have emerged. The first, AI ethics, aims to prescribe the best ethical principles for AI to follow. The second, AI value alignment, aims to align AI’s actions with humanity’s (or some culture’s) committed values. These approaches, however, face significant challenges. A viable alternative path between the AI ethics / value alignment dilemma focuses on human needs. The needs approach does not centralize ethical principles but it is consistent with them; and it does not aim for comprehensive value alignment because it focuses on our fundamental needs. Building on recent work by Montag, et al., I propose that AI should be designed around Maslow's hierarchy of needs. Beyond the standard Maslowian needs, I emphasize Maslowian preconditions – e.g., freedom of expression, freedom to reasonably act, freedom to investigate – required to fulfill further needs. These preconditions are best supported in cooperative communities. So, from an evolutionary point of view, I propose a three-stage process by which humans have developed cooperative communities. AI, I contend, can be integrated into and support these communities. By focusing on Maslowian needs and preconditions, AI will empower rather than disempower humans.