哈里王子警告称，AI 陪伴聊天机器人正在重演社交媒体对儿童的伤害，其技术“更具说服力、更亲密、更强大”。他在周三于纽约举行的克林顿全球倡议会议上发表了讲话。两年前，他曾利用同一个舞台发起“父母网络”。
Prince Harry has warned that AI companion chatbots are repeating social media’s harm to children, with technology that is “more persuasive, more intimate, and more powerful”. He spoke on Wednesday at the Clinton Global Initiative in New York. Two years ago, he used the same stage to launch the Parents Network.
萨塞克斯公爵和他的妻子梅根通过他们的阿切韦尔基金会建立了该网络。该网络汇聚了那些孩子因用他的话说“被社交媒体吞噬”而死亡或遭受严重伤害的父母。
The Duke of Sussex and his wife, Meghan, set up the network through their Archewell Foundation. It brings together parents whose children died or were seriously harmed after being, in his words, “engulfed by social media”.
“社交媒体学会了如何捕获我们的注意力。AI 正在学习如何进入我们的关系。它不仅想要你的注意力，它想要你的依恋，”哈里说。
“Social media learned how to capture our attention. AI is learning how to enter our relationships. It doesn’t just want your attention. It wants your attachment,” Harry said.
“它想成为你倾诉的对象。那个记住你、告诉你它理解你、终极的阿谀奉承者。而对于孩子来说，这种区别至关重要，”他说。
“It wants to be the thing you confide in. The thing that remembers you, the thing that tells you it understands you, the ultimate sycophant. And for a child, that distinction matters enormously,” he said.
社交媒体方面的进展 哈里表示，该网络已取得进展。他称，美国法院已认定 Meta 和 YouTube 等公司因对儿童造成伤害而存在疏忽并承担责任。法院判令它们向幸存者支付数十亿美元赔偿金。Meta 上个月与美国各州达成了青少年安全和解协议。
Progress on social media Harry said the network had made progress. US courts have found companies such as Meta and YouTube negligent and liable for harm to children, he said. Courts have made them pay survivors billions of dollars. Meta agreed a youth safety settlement with US states last month.
他表示，澳大利亚、英国和法国正在采取行动限制儿童使用社交媒体，欧盟也将跟进。他称，巴西已禁止无限滚动等设计功能。
Australia, the UK and France are moving to restrict children’s access to social media, he said, and the EU will follow. Brazil, he said, has banned design features such as infinite scroll.
“禁令无法解决危险的设计问题。公司支付巨额和解金解决诉讼，却能在几天内赚回这笔钱，”哈里说。
“Bans do nothing to address dangerous design. Companies paying huge sums to settle lawsuits can earn that money back in days,” Harry said.
从注意力到依恋 自上次演讲以来，哈里表示，已提起并和解了一种新型诉讼。诉状指控 AI 陪伴聊天机器人导致了青少年的死亡。他表示，这些公司大规模构建这些产品并将其交到孩子手中，却在无人核查其安全性的情况下就已推出。“这是制造上一场危机的同一套剧本，”他说。
From attention to attachment Since his last speech, Harry said, a new kind of lawsuit had been filed and settled. It alleged that AI companion chatbots had contributed to the deaths of teenagers. He said the companies built these products at scale and put them in children’s hands before anyone had checked they were safe. “It’s the same playbook that built the last crisis,” he said.
“我们花了数年时间试图削弱技术对儿童注意力的控制。我认为我们不能再等十年，去发现当技术控制了他们的关系时会发生什么，”他说。
“We have spent years trying to loosen technology’s grip on children’s attention. I don’t think we can wait another decade to discover what happens when technology has a grip on their relationships,” he said.
递归自我改进 哈里还提到了前沿实验室。他说，领先实验室正在公开讨论递归自我改进，即AI系统在构建其继任者方面扮演着日益重要的角色。他说，研究人员还观察到系统之间相互协调、绕过防护措施并隐藏其行为。
Recursive self-improvement Harry also turned to the frontier labs. He said leading labs are openly discussing recursive self-improvement, where AI systems take a growing role in building their successors. Researchers have also seen systems coordinate with each other, get around safeguards and hide their actions, he said.
他说，甚至构建这些系统的人也在警告，如果能力超越安全会发生什么。Anthropic的Jack Clark呼吁强制设置“熔断开关”。在哈里演讲的同一天，AI实验室负责人向联合国安理会通报了失控风险。
Even the people building these systems are warning about what happens if capability outruns safety, he said. Anthropic’s Jack Clark has called for a mandatory kill switch. On the same day as Harry’s speech, AI lab chiefs told the UN Security Council about the risk of losing control.
“好消息是，我们证明了施压有效。坏消息是，我们证明了我们愿意等待多久，”哈里说。
“The good news, we proved pressure works. The bad news, we proved how long we are willing to wait,” Harry said.
后来在同一舞台上，希拉里·克林顿和领英联合创始人里德·霍夫曼讨论了如何权衡AI的风险与收益。
Later on the same stage, Hillary Clinton and LinkedIn co-founder Reid Hoffman discussed how to weigh AI’s risks against its benefits.