Automated Moderation Is Here to Stay (2-part series)

This EFF series revisits the organization's 2020 position on automated content moderation, concluding that AI-driven moderation has shifted from a temporary, crisis-era measure into permanent infrastructure across major platforms. Rather than original data collection, the piece synthesizes findings from bodies including the UN, OSCE, OAS, the African Commission on Human and Peoples' Rights, Human Rights Watch, and the Center for Democracy and Technology, tracing the shift back to Facebook's 2017-2018 adoption of AI flagging (at one point covering 99% of removed ISIS and Al Qaeda content) and the pandemic-driven reduction of human review capacity in 2020.

The authors identify recurring harms: over-removal and discriminatory enforcement stemming from biased training data, disproportionate impact on marginalized communities and speakers of lower-resource languages, and erasure of documentation with evidentiary value, including human rights and war-crimes material.

They conclude that the relevant policy question is no longer whether platforms will rely on automated moderation but under what conditions they should, and call for stronger transparency, accountability, and appeals safeguards to keep pace with the technology's now-permanent role.