2025年10月, AI生成的飓风视频误导了社交媒体用户, AI-generated hurricane videos misled social media users in Oct 2025, spreading fear and confusion despite watermarks.

当2025年10月底梅利萨飓风接近牙买加时, AI制作的影片在TikTok、Facebook和Instagram上迅速传播, 描述风暴破坏、洪水和人类危难的假景景象。 As Hurricane Melissa neared Jamaica in late October 2025, AI-generated videos depicting false scenes of storm damage, flooding, and human distress spread rapidly on TikTok, Facebook, and Instagram. 尽管OpenAI的Sora模型有水印, 但许多观众仍不承认内容是合成的, 导致情绪反应和错误信息。 Despite watermarks from OpenAI’s Sora model, many viewers failed to recognize the content as synthetic, leading to emotional reactions and misinformation. 包括新闻部长Dana Morris Dixon在内的官员敦促依赖经核实的来源,警告假冒可能破坏应急准备。 Officials, including Information Minister Dana Morris Dixon, urged reliance on verified sources, warning the fakes could undermine emergency preparedness. Hany Farid和Amy McGovern等专家强调,在危机期间,AI驱动的欺骗威胁日益增大,他们指出,甚至贴有标签的内容也常常无人注意,增加了公众的困惑和风险。 Experts like Hany Farid and Amy McGovern highlighted the growing threat of AI-driven deception during crises, noting that even labeled content often goes unnoticed, increasing public confusion and risk. 要求AI标签的Meta政策似乎执行不一贯, Meta’s policies requiring AI labels appear inconsistently enforced, raising concerns about trust and safety during high-stakes events.