A recent study by the ADL’s Center for Technology and Society reveals alarming findings regarding AI video generators, which produced antisemitic or extremist content in 40% of trials using prompts across four platforms: Google’s Veo 3, OpenAI’s Sora 1 and Sora 2, and Hedra’s Character-3. OpenAI’s Sora 2 model showed the most promise, filtering out 60% of harmful prompts, while the other systems performed poorly, with Sora 1 refusing none. The findings highlighted the propagation of antisemitic tropes and conspiracy theories, including mythical imagery like a “Jewish man drinking blood.” Additionally, these tools generated violent content linked to known hate groups. The ADL emphasizes the urgent need for tech companies to enhance content moderation and training to prevent the dissemination of hate and disinformation. Stronger regulations and disclosure requirements for AI-generated content were also recommended to combat the recruitment of youth into extremist ideologies.
Source link
