This paper examines the relationship between libraries, artificial intelligence and the common good from the perspectives of inclusion, sustainability and social responsibility. Moving from the MIT AI Risk Repository, it focuses on three areas of risk that are especially relevant to information professions: discrimination and toxicity, misinformation and disinformation, and socioeconomic and environmental harms. The paper shows how AI systems may reproduce and amplify existing inequalities, particularly gender and intersectional biases, while also contributing to the degradation of information ecosystems through misleading, decontextualized or manipulated content. It further highlights the material and social infrastructures behind AI, including the invisible labour of data workers and the growing environmental impact of data centers. Against this background, libraries are presented not merely as points of access to new technologies, but as spaces of critical mediation, public education and democratic accountability. The paper argues that AI literacy should be understood as an expanded dimension of information literacy, capable of integrating ethical, social, ecological and political awareness. Only by recognizing both the potential and the risks of AI can libraries help orient algorithmic systems toward inclusive, sustainable and genuinely common-good objectives.