OpenAI叫停Sora马丁·路德·金视频创作,因用户制作“不敬”深度伪造

OpenAI叫停Sora马丁·路德·金视频创作,因用户制作“不敬”深度伪造

2025-10-21Technology
--:--
--:--
雷总
小王,晚上好,我是雷总,欢迎收听专属于你的 Goose Pod。今天是10月21日,星期二,晚上9点22分。
董小姐
我是董小姐。今晚我们聊一个热门话题:OpenAI 紧急叫停了用其视频生成工具 Sora 创作马丁·路德·金的视频。
雷总
我们开门见山吧。董小姐,这件事其实很有代表性。Sora 这个产品一发布,大家都在秀各种创意视频,结果有人用它生成了马丁·路德·金的形象,而且是一些“不敬”的描绘,这下就踩到红线了。
董小姐
没错。马丁·路德·金的家人和遗产委员会直接提出了抗议。OpenAI的反应也很快,立刻暂停了这个功能。他们说,公众人物及其家人,理应有权控制自己形象的使用方式。这一点我非常赞同,个人品牌和形象不容侵犯。
雷总
是的,这恰恰说明了现在AI模型的能力已经非常强大,可以轻松生成以假乱真的视频。技术跑得太快,规则就必须得跟上。我们做产品的,必须把这种风险控制住,这是一个技术问题,更是一个责任问题。
董小姐
你说到点子上了。能力越强,责任越大。技术本身没有对错,但使用技术的人,和提供技术的企业,必须要有底线和原则。
董小姐
其实,这已经不是第一次了。Sora 发布后,乱象就没断过。前段时间,已故喜剧演员罗宾·威廉姆斯的女儿就公开请求大家,不要再用AI生成他父亲的视频了,她说那些视频让她感到“不安”。
雷总
对,我看到了。所以有AI伦理学家批评OpenAI,说他们的策略是“用消防水龙带救火式的试错”,先把功能放出来,等出了问题再去补救。从产品迭代的角度看,快速试错有它的道理,但这次,确实是把双刃剑。
董小姐
我不同意这种“先污染,后治理”的逻辑。核心科技,不仅要技术过硬,更要安全可控。一个好的产品,不应该给社会带来这么大的伦理风险。难道发布之前,他们没有预见到这些问题吗?
雷总
他们肯定预见到了,所以也推出了一些所谓的“护栏”。比如,OpenAI现在的政策是,只要有授权的代表或遗产所有者提出要求,Sora就不能使用其形象。这算是他们给出的一个解决方案,一个技术上的补丁。
董小姐
这个补丁就引出了另一个问题:公平性。是不是只有名人才能获得这种“数字复活”的保护?我们普通人去世后,我们的形象是不是就可以被任何人随意使用了?这不合理。
雷总
这个问题非常尖锐。而且,深伪技术还带来一个更隐蔽的危险,叫做“骗子红利”。就是说,因为大家都知道有深伪技术存在,当某个公众人物真的爆出丑闻时,他就可以轻易地辩称“那是AI伪造的”,从而逃避责任。
董小姐
这简直是为说谎者提供了完美的借口!它会从根本上侵蚀我们社会对事实和证据的信任。这对商业、政治,甚至整个社会秩序的冲击,都是灾难性的。企业的信誉,政治家的承诺,都可能因此变得一文不值。
雷总
是的,这就是技术发展和社会伦理之间的巨大冲突。一方面是言论自由和创作自由,另一方面是个人权利和公共信任。AI在做内容审核时,很难理解人类复杂的语境和意图,要么过度审查,要么就让虚假信息泛滥。
董小姐
所以,不能把宝全押在技术上。平台的责任、法律的界定,缺一不可。技术必须为人服务,而不是制造混乱。
雷总
现在最大的影响,就是公众对“眼见为实”这个词产生了巨大的怀疑。一段视频,一张图片,我们都需要先在心里打个问号。OpenAI这次的举动,虽然是被动的,但也是在尝试为“同意”和“可追溯”设定一个行业先例。
董小姐
一个问号可不够。以后可能任何影像证据都可能被挑战。这对司法、新闻等依赖视觉证据的领域是致命的。我听说OpenAI给视频加上了水印,这个措施能起多大作用?
雷总
是的,他们用了可见水印和C2PA元数据两种方式来标记AI生成的内容。这就像给数字内容办了个“身份证”。但这更像是一场“军备竞赛”,总有技术可以绕过这些标记。不过,这是建立信任必须走出的第一步。
雷总
展望未来,我相信技术的问题,最终还是要靠技术来解决。比如,用更强大的AI来检测AI生成的内容,甚至可以利用区块链技术来验证内容的原始出处和真实性。这听起来是不是很酷?
董小姐
听起来很理想,但光有技术还不够。法律法规必须迅速跟上,为技术划定明确的红线。同时,企业必须承担起社会责任,将伦理规范融入产品设计中。这才是企业真正的、可持续的核心竞争力。
雷总
说得对。好了,今天的讨论就到这里。感谢小王收听Goose Pod。
董小姐
我们明天再见。

Here is a comprehensive summary of the news article: **News Summary: OpenAI Halts Sora Videos of Martin Luther King Jr. After 'Disrespectful' Deepfakes** * **News Provider:** CNBC * **Author:** Samantha Subin * **Publication Date:** October 17, 2025 --- **Key Findings & Developments:** * **Halt on MLK Jr. Videos:** OpenAI has suspended the generation of artificial intelligence-generated videos depicting Dr. Martin Luther King Jr. using its short-form video tool, **Sora**. * **Reason for Halt:** The decision was made after users employed Sora to create **"disrespectful depictions"** of the civil rights leader. * **OpenAI's Policy Shift:** * OpenAI stated that while **free speech interests** are considered in depicting historical figures, public figures and their families should ultimately have **control over how their likeness is used**. * The company plans to implement tougher **"guardrails"** for generating content involving historical figures. * Public figures or their representatives will have the ability to **request not to appear** in Sora videos. * **Sora Tool Performance:** * Sora, which allows users to create AI-generated short videos from text prompts, was launched at the **end of September**. * In **less than five days**, Sora amassed over **1 million downloads**, achieving this milestone faster than ChatGPT. **Broader Context and Related Concerns:** * **Precedent of Deepfake Issues:** This incident is part of a growing trend of public figures raising concerns about AI-generated deepfakes: * **Zelda Williams**, daughter of the late comedian Robin Williams, has requested that people cease sending her AI videos of her father. * Last year, actress **Scarlett Johansson** reported that OpenAI used a voice on ChatGPT that sounded **"eerily similar"** to her performance in the movie "Her." OpenAI subsequently removed the voice from its platform. * **Risks Associated with AI-Generated Videos:** The rapid rise of AI video generation tools like Sora raises significant concerns, including: * The spread of **misinformation**. * **Copyright infringement**. * The proliferation of **"AI slop"** – quickly produced videos that flood social media feeds. **Material Financial Data Mentioned:** * The article includes a reference to **"$1 trillion-worth of AI deals between OpenAI, Nvidia and others"** as a related piece of tech news, presented in a chart. This figure represents the scale of investment and activity in the AI sector but is not directly tied to the Sora/MLK Jr. incident itself.

OpenAI stops Sora videos of Martin Luther King Jr. after users made ‘disrespectful’ deepfakes

Read original at CNBC

Dr. Martin Luther King Jr. addressing crowd of demonstrators outside the Lincoln Memorial during the March on Washington for Jobs and Freedom.Francis Miller/The LIFE Picture Collection via Getty ImagesOpenAI halted artificial intelligence-generated videos of Martin Luther King Jr. after users utilized its short-form video tool Sora to create "disrespectful depictions" of the civil rights leader."

While there are strong free speech interests in depicting historical figures, OpenAI believes public figures and their families should ultimately have control over how their likeness is used," OpenAI said in a post to social media platform X.The ChatGPT maker said it will work to toughen "guardrails" for historical figures and that public figures or representatives can ask to not appear in Sora videos.

OpenAI did not immediately respond to CNBC's request for comment.Read more CNBC tech news$1 trillion-worth of AI deals between OpenAI, Nvidia and others, in one chartWaymo plans robotaxi launch in London, marking its European debutASML looks to calm fears over 2026 growth as it warns of China sales declineMeta removes Facebook page allegedly used to target ICE agents after pressure from DOJOther public figures have also called out the use of AI deepfakes.

Last week, Zelda Williams, the daughter of late comedian Robin Williams, asked that people stop sending her AI videos of her father.Last year, actress Scarlett Johansson said the company used a voice that sounded "eerily similar" to her performance in the movie "Her" on ChatGPT. OpenAI later pulled the voice from its platform.

OpenAI launched Sora at the end of September. The tool allows users to create AI-generated short videos using a text prompt. In less than five days, Sora head Bill Peebles said the tool had amassed over 1 million downloads, hitting the milestone faster than ChatGPT.Its ascent and the rise of AI-generated videos have also raised questions and concerns over the spread of misinformation, copyright infringement and the proliferation of AI slop, quickly produced videos that flood social feeds.

Analysis

Conflict+
Core Event+
Background+
Impact+
Future+

Related Podcasts

OpenAI叫停Sora马丁·路德·金视频创作,因用户制作“不敬”深度伪造 | Goose Pod | Goose Pod