awareness, they will conflict with an organization’s core values of promoting the public interest and human rights, and may undermine trust in the organization. For this reason, internal procedures are necessary to detect and prevent these risks when using generative AI. To prevent these risks, it is first necessary to provide regular human rights training so that all activists are able to recognize bias and discrimination in AI-generated outputs. Depending on the organization’s needs, it may also be appropriate to establish procedures—or designate responsible reviewers—to conduct prior review of all materials intended for external publication. If outputs are suspected of containing hateful or discriminatory expressions, their use should be immediately suspended and the matter referred to the designated reviewer. Alternatively, the organization may request revisions from the generative AI system itself (e.g., “This expression may be discriminatory; please rewrite it using neutral and inclusive language”). The revised output should then be reviewed again to ensure that no problematic expressions remain. If the AI system produces seriously problematic content, or repeatedly generates discriminatory or hateful expressions, the organization should raise the issue through the AI provider’s reporting or feedback channels. If the same problems recur or are not adequately addressed, the organization should formally discontinue use of the service. Alternative tools should then be considered, and the problematic cases should be documented internally to help prevent recurrence. Generative AI Guide for Civil Society