When **Frances Haugen** blew the whistle on **Meta** in 2020, a stark reality emerged: the company's algorithms, designed to detect terrorist content, incorrectly deleted nonviolent Arabic-language content a staggering 77 percent of the time. Simultaneously, these systems frequently failed to identify hate speech, a problem **Meta's** own transparency report later corroborated. Five years on, researchers in the region report that overzealous moderation persists, with avenues for redress largely inaccessible. This issue is amplified in less-resourced languages, where the problem escalates from faltering to outright failure. ## The Language Barrier and Systemic Bias A 2025 report from the **Center for Democracy and Technology** revealed inconsistencies, biases, and inaccuracies in labeled datasets for languages and dialects such as Maghrebi Arabic and Kiswahili. This stems from limited hiring of annotators who are native speakers and a failure to adapt to linguistic shifts. An investigation into **ChatGPT's** performance in several low-resource languages further underscores the depth of this challenge. Language disparities are just one facet of the concerns surrounding widespread automated moderation. From the systemic suppression of content related to Palestine to the repeated misclassification of LGBTQ+ content as explicit material, these examples highlight the risks of over-reliance on automated systems and the urgent need for stronger safeguards. ## Balancing Scale with Human Rights Automated systems can process content at a scale unattainable by humans, potentially easing the psychological burden on human moderators. However, these systems inherently reproduce existing biases, struggle with contextual understanding, and often make mistakes that disproportionately affect journalists, activists, artists, and other vulnerable communities. As Rachel Griffin articulated in 2023, "Perfectly accurate moderation is not only technically out of reach but intrinsically impossible." Despite these inherent flaws, companies, policymakers, and civil society can take significant steps to ensure highly automated systems respect human rights, minimize harm, and provide accountability when errors occur. If companies continue to depend on automation for content moderation, accountability frameworks must evolve in parallel with these technologies. ## The Path Forward: Transparency, Cultural Competence, and Appeals This evolution can begin with a commitment to the **Santa Clara Principles 2.0**. These principles, updated in 2021, reflect global community needs and specifically address automation. The first Foundational Principle states: > Companies should ensure that human rights and due process considerations are integrated at all stages of the content moderation process, and should publish information outlining how this integration is made. Companies should only use automated processes to identify or remove content or suspend accounts, whether supplemented by human review or not, when there is sufficiently high confidence in the quality and accuracy of those processes. Companies should also provide users with clear and accessible methods of obtaining support in the event of content and account action. Drawing on the **Santa Clara Principles 2.0**, international human rights standards, and years of research documenting the shortcomings of automated moderation, we propose eight key recommendations for policymakers and companies deploying AI-assisted content moderation systems: 1. Automated technologies should augment, not replace, human moderators. Systems can flag and prioritize content, while humans interpret context, handle sensitive cases, and refine system performance. 2. Companies must be transparent about the specific instances and methods of automation used in content decisions. 3. Regular audits of automated systems for bias are crucial, with particular attention to low-resource languages, vulnerable communities, and conflict zones. 4. Users must have accessible avenues to appeal moderation decisions, providing context when they believe content has been wrongfully removed. Appeals should be promptly evaluated by human moderators. 5. Companies should regularly assess the human rights impact of their moderation decisions and publicly report the results. 6. If third-party vendors are utilized, companies must carefully and regularly audit them for compliance with these principles. 7. Lawmakers should avoid legislation that explicitly or effectively mandates automated moderation systems. 8. Policymakers should refrain from dictating platforms' technical and design choices to favor or disfavor specific expressions. These recommendations underscore that automated content moderation is not merely a technical challenge. Given its profound impact on public discourse and fundamental rights, its design and oversight must actively incorporate the concerns of policymakers, civil society, independent researchers, and the communities most affected by these powerful systems.