A/80/496 during a campaign of ethnic violence, which has had horrific and long -lasting humanitarian consequences. Not only did Facebook fail to effectively moderate the hateful content, but also its profit-driven content-shaping algorithms have been implicated in the amplification of racist hate speech against the Rohingya ethnic group. The relentless profit-seeking reportedly continued, notwithstanding multiple warnings of grave impending harm. 23 18. The development of generative artificial intelligence may intensify online racist hate speech, including incitement, and its role in driving conflict and violence. Generative artificial intelligence systems are newer forms of artificial intelligence that can produce a wide range of output, including images, text, audio, video and synthetic data. As described above, hate speech and misinformation and disinformation are deeply interconnected. Incorrect information, in particular racial stereotyping, distortions of historical atrocities against groups and the scapegoating of those from marginalized racial and ethnic groups, spread sometimes unknowingly and often intentionally, can stroke racial hatred and contribute to violent conflict. The potential of generative artificial intelligence to produce racially discriminatory “deep fakes” was included in information received by the Special Rapporteur. 24 19. The Special Rapporteur also received information about how the moderation of online content, which often involves algorithms, can censor the voices of those from marginalized racial and ethnic groups, including content relating to conflict situations. This was described in one submission as an “automated censorship system that reproduces structural inequalities”. 25 Posts by Indigenous Peoples, people of African descent, women and LGTBIQI+ people, are reportedly frequently misinterpreted or deleted by algorithms that identify terms such as “racism” or “gender violence” as potentially dangerous. The Special Rapporteur is also concerned by reports of the disproportionate censorship of Palestinian content on Meta’s Facebook and Instagram platforms. 26 The combination of a failure to address racial hate speech, even in cases that amount to incitement to discrimination, hostility and violence, and restrictions on the freedom of expression of those from marginalized racial and ethnic groups creates online marginalization and reproduces harmful racial hierarchies. On the other hand, there is also potential for the positive application of artificial intelligence to the moderation of online content if it is carefully designed and deployed to avoid algorithmic bias further reinforcing harmful racial hierarchies. For example, the Special Rapporteur receives information about promising research demonstrating that artificial intelligence tools can identify online antisemitic racist hate speech. 27 20. There are other potentially positive applications of artificial intelligence to conflict prevention and resolution. For example, artificial intelligence is starting to be used to optimize the delivery of humanitarian aid and could play a positive role in improving early warning systems. 28 While the Special Rapporteur welcomes initiatives that deploy emerging technologies to help to prevent conflict or address its racialized impact, she is concerned about the deployment of such technologies in such a high-risk setting. Even in cases in which the goals of the application of artificial intelligence are laudable, the largely unregulated development of artificial intelligence and a tendency of any management initiatives to be “race-neutral” means there is a risk of algorithmic bias, resulting in dangerous outcomes for those from marginalized __________________ 23 24 25 26 27 28 8/21 See A/HRC/46/57. Submission from World Jewish Congress. Submission from FDI. Submission from the State of Palestine. See also, Human Rights Watch, “Meta’s broken promises: systemic censorship of Palestine content on Instagram and Facebook ”, 21 December 2023. Submission from World Jewish Congress. Submission from South Africa. 25-16958