Our communication on the Internet is increasingly being shaped by Artificial Intelligence (AI) – both as a driver of harmful online content and as a critical tool for its mitigation. Generative AI systems now enable the large-scale creation of persuasive hate speech and disinformation, while algorithmic moderation and detection tools raise pressing concerns about bias, accountability, and freedom of expression. This Dagstuhl Seminar aims to bring together a cohort of leading researchers from computer science, law, and the social sciences to examine how societies can effectively combat online hate speech in the era of AI-driven communication. Participants will explore technical challenges including multimodal detection, the scaling of AI for long-tail languages, and explainable moderation alongside the legal hurdles of attributing criminal accountability for AI-generated abuse across diverging jurisdictions. Special attention will be given to regulatory gaps arising from insufficient consideration of multilingual and cultural variation in increasingly opaque digital spaces. Through interdisciplinary dialogue, the seminar will develop shared principles for responsible AI and propose pathways toward transparent, fair, and context-sensitive moderation systems. The deliverables will emerge through discussion organized around these broad research challenges:

Technical Challenges: How can LLMs and generative models be harnessed to detect multi-modal hate speech? How to scale AI detection on long-tail of languages? How can we detect hate-speech generated by AI? How can we make content moderation decisions made by AI more explainable?

Legal Challenges: How does criminal liability for online hate speech vary in different jurisdictions? What is the legal accountability of online hate speech generated by AI? How can hate speech generated by AI be prosecuted according to criminal law? What are the challenges in attributing criminal liability to AI systems?

The primary aim of the seminar is to bridge the gap between technical innovation and legal accountability in the context of AI-driven communication. To this end, the goals and outcomes are consolidated into two deliverables aligned with the identified topics.

Joint Strategic Roadmap: To ensure long-term impact, the seminar will produce a roadmap outlining the most urgent open technical, legal, and regulatory hurdles in combating online hate speech in the age of AI. We aim to formalise them through a joint publication, fostering a sustained interdisciplinary network of researchers and practitioners.

Legal Accountability Framework: A policy brief detailing pathways for attributing criminal liability to AI systems that generate synthetic online hate speech. This outcome will focus on the practical difficulties in attributing criminal liability to automated systems, and enforcing laws across digital borders, ensuring these legal frameworks are technically feasible for auditing and prosecution.

Creative Commons BY 4.0

Vaibhav Bajpai, Eva Richter, Caitlin Carlson, Susan Benesch, Jon Crowcroft, and Sonal Khosla