42 No. 78 JUSTICE persuasive capacity means that, beyond self-harm, there is potential for radicalization or reinforcement of prejudiced ideologies, including antisemitism and other forms of racial hatred or extremist ideologies. Emotional engagement increases susceptibility to harmful narratives, and without proper safeguards, AI companions may unwittingly or deliberately amplify them. Recognizing some of the risks, the state of California recently introduced a law targeting “companion chatbots,” that includes guardrails against persuading users to self-harm and an obligation to remind users that they are conversing with a machine.24 The harms associated with AI‑generated antisemitism are both direct and systemic. At an individual level, victims may face harassment, reputational damage, or credible threats based on fabricated content. At a communal level, Jewish communities experience heightened fear, stigmatization, and marginalization. More broadly, AI‑enabled antisemitism undermines democratic discourse. Deepfakes and synthetic narratives erode trust in information ecosystems, making it harder for the public to distinguish truth from fabrication. This erosion of epistemic trust disproportionately harms minority communities, which are more likely to be targeted by disinformation campaigns. From a regulatory perspective, AI systems introduce challenges of scale, speed, and attribution. Harmful content can be generated faster than human moderators can respond, and responsibility may be diffused across users, developers, deployers, and platforms. These dynamics strain regulatory models that were designed for human‑authored content. The case for regulating AI in the context of antisemitism is not rooted in abstract technological pessimism, but in documented patterns of harm. Existing online safety frameworks, including those in the UK, largely assume that harmful content is created by humans and merely hosted or amplified by platforms. AI disrupts this assumption by enabling the automated production of harmful material at unprecedented scale. UK Online Safety and AI Regulation: Addressing Antisemitic Harm The Online Safety Act The UK’s Online Safety Act 2023 (OSA) regulates online content. Rather than creating new speech offenses, the Act imposes systems and process duties on regulated services, including user-to-user platforms and search engines. Ofcom is designated as the independent regulator, empowered to issue codes of practice, require risk assessments, and impose significant financial penalties for non-compliance. For the purposes of antisemitism, the Act is relevant in three principal ways: its treatment of illegal content, its duties relating to content harmful to adults and to children, and its broader emphasis on risk assessment and mitigation. Importantly, the OSA does not regulate content directly; instead, it regulates how platforms manage and reduce the risk of harm arising from content. Antisemitic content that meets the threshold of criminality, such as threats of violence, harassment, or incitement under existing hate crime legislation, falls squarely within the OSA’s illegal content regime. Platforms of all sizes are required to have systems and processes to address illegal antisemitic content once they are made aware of it. However, proactive duties such as systemic risk assessments, auditing, and enhanced safety-by-design requirements are applied more extensively to large and designated services. In theory, this framework provides a robust mechanism for addressing the most egregious forms of online antisemitism. In practice, however, there has been persistent under-enforcement and inconsistent application across platforms. Much antisemitic content remains online for extended periods, particularly when it is framed as political commentary or coded speech rather than explicit racial hatred.25 So too, there will always be a degree of difficulty in judging where content crosses the threshold, the “borderline” into illegality. There will always be a risk that platforms make differing – and sometimes incorrect – judgments about whether content should remain online. 24. Stuart D. Levi, Michael W. McTigue, Jr. and Meredith C. Slawe, “New California ‘Companion Chatbot’ Law: Imposes Disclosure, Safety Protocol and Annual Reporting Requirements” (Oct. 17, 2025), available at https://www.skadden. com/insights/publications/2025/10/new-california-companion-chatbot-law 25. C. Sherwood, “New online laws fail to stop promotion of terror groups” (June 26, 2025), available at https://www. thetimes.com/uk/crime/article/new-online-laws-fail-to-stop-promotion-of-terror-groups-m92g2fcxb
RkJQdWJsaXNoZXIy MjgzNzA=