AI Content Moderation Policy

AI content moderation policy is the discipline of deciding what stays up on a platform, what comes down, what gets labelled, and how those decisions are explained — and proving that the system is consistent, lawful, and humane. It sits at the intersection of platform-policy writing, applied ethics, intermediary-liability law (Section 230, the EU DSA, the UK Online Safety Act, NetzDG, India IT Rules), the trust & safety operations stack (queues, reviewers, escalations, appeals), the AI / ML moderation stack (toxicity / hate / NSFW classifiers, LLM moderation, hashing, multimodal), and the transparency machinery that runs in production (statements of reasons, transparency reports, oversight boards, regulator inspections). Over the last five years it has stopped being a back-office function and has become an operating commitment subject to direct regulatory duties, multi-million-euro fines, civil-society scrutiny, and journalist investigation.

This track is written for the practitioners doing this work day to day: trust & safety policy leads, T&S operations leaders, ML engineers building moderation systems, integrity / civic teams, appeals and oversight-liaison teams, lawyers translating intermediary-liability law into operational controls, and program leads stitching the program together. Every topic explains the underlying moderation discipline (drawing on Gillespie, Roberts, Klonick, Suzor, Douek and the canonical T&S literature, Santa Clara Principles, the Christchurch Call, the DSA, and hard-won production experience), the practical artefacts and rituals that operationalise it (policy specs, reviewer guidance, runbooks, transparency reports, audit packets), and the failure modes where moderation work quietly breaks down in practice. The aim is that a reader can stand up a credible content-moderation function, integrate it with engineering and governance, and defend it to boards, regulators, journalists, oversight boards, and the users the system actually affects.