Moderating terrorist live-streaming presents legal and technical trade-offs. The immediacy of live-streamed content requires a timely assessment of the content’s compliance with community guidelines and an even more expeditious action of restriction or removal in case of portrayal of illegal content. Social media companies have heavily relied on technology to moderate content. The most frequently used tools to screen and filter content and ensure compliance with community standards and regulations include hashing technology, video-fingerprinting, natural language processing and metadata analysis. Terrorist content online presents nuances that make the task of removing it particularly challenging especially when companies must rely on machine learning content moderation. We identify three trade-offs in regulating content moderation of live-streamed content: technological neutrality v specificity, explainability v adversarial machine learning, and accuracy v time-efficiency. For each trade-off, we provide an analysis of the considerations lawmakers and regulators must take into account when balancing competing interests.

中文翻译：

---

恐怖主义直播内容审核的法律和技术权衡

缓和恐怖主义直播需要在法律和技术上进行权衡。直播内容的即时性要求及时评估内容是否符合社区准则，并在描绘非法内容的情况下采取更迅速的限制或删除行动。社交媒体公司严重依赖技术来管理内容。用于筛选和过滤内容并确保遵守社区标准和法规的最常用工具包括散列技术、视频指纹识别、自然语言处理和元数​​据分析。在线恐怖主义内容呈现出细微差别，这使得删除它的任务特别具有挑战性，尤其是当公司必须依赖机器学习内容审核时。我们确定了在规范直播内容的内容审核方面的三个权衡：技术中立性与特异性、可解释性与对抗性机器学习以及准确性与时间效率。对于每项权衡，我们分析了立法者和监管者在平衡竞争利益时必须考虑的因素。