OpenAI首席执行官山姆·奥尔特曼（Sam Altman）宣布，ChatGPT即将面向通过年龄验证的成年用户开放情色内容生成功能。奥尔特曼将此举阐释为“像对待成年人一样对待成年用户”。消息发布之际，正值所谓的“AI精神错乱”案例增多，OpenAI因其AI聊天机器人可能影响用户心理健康而面临审查。此前，包括埃隆·马斯克（Elon Musk）旗下的xAI在内的竞争对手，已在其平台上推出了性暗示更明显的聊天机器人“伴侣”。
市场对能够进行浪漫互动的AI聊天机器人显然存在巨大需求。《哈佛商业评论》（Harvard Business Review）今年4月对6000名常规AI用户的调查发现，“陪伴和治疗”是最常见的用途。另据方舟投资（Ark Invest）的一项研究，专注于成人内容的AI平台去年取得了显著增长，市场份额从前一年的仅1.5%大幅提升至14.5%，抢占了此前由OnlyFans主导的市场。
一些以AI聊天伴侣为卖点的流行替代品，如Character.ai（Character.ai）和Replika（Replika），也印证了用户对此日益增长的需求。今年早些时候，xAI为其聊天机器人Grok推出了“伴侣模式”，该功能允许用户与各种角色互动，其中包括一个高度性化的动漫角色“Ani”。
乔治城大学安全与新兴技术中心（Center for Security and Emerging Technology）的高级研究分析师杰西卡·李（Jessica Li）告诉《财富》杂志：“OpenAI有点陷入进退两难的境地，因为他们看到了来自用户的强烈需求信号。就用户与模型建立的关系而言……情色内容或成人内容也属于这种与模型情感互动的范畴。”
Replika.李表示，此举可能是OpenAI试图“走钢丝”：一方面通过承诺对不适宜工作场所（NSFW）内容提供可选的内容审核来维持现有市场份额，另一方面也试探是否能从Replika等更专业或小众的服务中争夺用户。
李补充道：“尽管有一些关于打造能提振经济的通用人工智能的叙事，但OpenAI仍然试图作为一个技术平台运营，某种程度上也像一家社交媒体公司。在兜售给投资者和政客的叙事……与市场上实际发生的事情之间，存在着一种有趣的张力。”
OpenAI涉足成人内容领域的举动，招致了儿童安全倡导者和业内知名人士的批评，他们担心即使有年龄验证，情色内容仍可能触及年轻用户。
本周早些时候，企业家兼电视名人马克·库班（Mark Cuban）表示，OpenAI的计划可能“引发强烈反效果”，并认为父母不会相信OpenAI的年龄过滤器能让孩子远离露骨内容。在美国，联邦贸易委员会（Federal Trade Commission）已就AI聊天机器人与未成年人的互动方式展开调查，各州立法者也在考虑加强对数字伴侣和性化AI内容的监管。Boies Schiller Flexner律师事务所的合伙人珍妮·金（Jenny Kim）告诉BBC，OpenAI正在“把人们当作小白鼠”，并质疑该公司将如何防止儿童在平台上访问成人内容。
OpenAI在回应置评请求时表示，公司正在构建一个年龄预测系统，以判断用户是否年满18岁。他们补充说，如果无法确认用户的年龄，聊天机器人将采取更安全的路径，默认提供18岁以下用户的体验，同时为成年人提供证明年龄以解锁成人功能的方式。
针对X平台上的一些反对声浪，奥尔特曼（Altman）回应称，这次宣布在“情色内容这一点上引发的反响远超他的预期”。他强调，这一变化“只是我们允许成年用户拥有更多自由的一个例子”，并非放弃安全措施或心理健康防护。
“我们不是由选举产生的世界道德警察，”他说，并补充道，ChatGPT将继续“优先考虑青少年的安全，而非其隐私和自由”，同时给予成年人更多自主权。
GPT-4o引发的反弹
OpenAI还一直在应对一波始料未及的反弹，此前的决定是用其更新的GPT-5模型取代由GPT-4o驱动的ChatGPT版本。用户们抵制这一变更，称他们失去了与旧版聊天机器人建立的AI友谊和浪漫关系。一份要求保留旧版聊天机器人的请愿书已收集到近六千个签名。
请愿者在change.org活动页面写道：“对我们许多人来说，GPT-4o提供了独特且不可替代的用户体验，它融合了我们所看重的品质和能力，这与性能基准无关。我们继续以独特而有意义的方式从GPT-4o中受益。”
尽管OpenAI最终恢复了旧版本，但参与#keep4o运动的人士此后告诉《财富》杂志，他们担心公司会“在未经同意或通知的情况下”将用户导向GPT-5。
据杰西卡·李分析，OpenAI的宣布可能试图向这些用户传递某种信息：“如此公开地宣布此事，确实让我觉得他们是在试图向那些有此需求的用户发出信号——比如，‘我们听到了你们的声音。我们正在回应你们的渴望。’”
这一转变也并非全新；它建立在今年早些时候对OpenAI模型规范（Model Spec）一次更低调的更新之上。今年2月，OpenAI更新了该文件，放宽了对性和暴力内容的限制，称此举是为了摆脱“AI家长式作风”。当时的更新指南允许在适当的语境下生成书面情色内容和其他敏感材料。OpenAI还告诉《财富》杂志，此次宣布是其基于最新发布的模型规范制定计划的一部分，同时会维持底线，抵制深度伪造等有害用途。（*）
译者：郝秀
审校：汪皓
OpenAI首席执行官山姆·奥尔特曼（Sam Altman）宣布，ChatGPT即将面向通过年龄验证的成年用户开放情色内容生成功能。奥尔特曼将此举阐释为“像对待成年人一样对待成年用户”。消息发布之际，正值所谓的“AI精神错乱”案例增多，OpenAI因其AI聊天机器人可能影响用户心理健康而面临审查。此前，包括埃隆·马斯克（Elon Musk）旗下的xAI在内的竞争对手，已在其平台上推出了性暗示更明显的聊天机器人“伴侣”。
市场对能够进行浪漫互动的AI聊天机器人显然存在巨大需求。《哈佛商业评论》（Harvard Business Review）今年4月对6000名常规AI用户的调查发现，“陪伴和治疗”是最常见的用途。另据方舟投资（Ark Invest）的一项研究，专注于成人内容的AI平台去年取得了显著增长，市场份额从前一年的仅1.5%大幅提升至14.5%，抢占了此前由OnlyFans主导的市场。
一些以AI聊天伴侣为卖点的流行替代品，如Character.ai（Character.ai）和Replika（Replika），也印证了用户对此日益增长的需求。今年早些时候，xAI为其聊天机器人Grok推出了“伴侣模式”，该功能允许用户与各种角色互动，其中包括一个高度性化的动漫角色“Ani”。
乔治城大学安全与新兴技术中心（Center for Security and Emerging Technology）的高级研究分析师杰西卡·李（Jessica Li）告诉《财富》杂志：“OpenAI有点陷入进退两难的境地，因为他们看到了来自用户的强烈需求信号。就用户与模型建立的关系而言……情色内容或成人内容也属于这种与模型情感互动的范畴。”
Replika.李表示，此举可能是OpenAI试图“走钢丝”：一方面通过承诺对不适宜工作场所（NSFW）内容提供可选的内容审核来维持现有市场份额，另一方面也试探是否能从Replika等更专业或小众的服务中争夺用户。
李补充道：“尽管有一些关于打造能提振经济的通用人工智能的叙事，但OpenAI仍然试图作为一个技术平台运营，某种程度上也像一家社交媒体公司。在兜售给投资者和政客的叙事……与市场上实际发生的事情之间，存在着一种有趣的张力。”
OpenAI涉足成人内容领域的举动，招致了儿童安全倡导者和业内知名人士的批评，他们担心即使有年龄验证，情色内容仍可能触及年轻用户。
本周早些时候，企业家兼电视名人马克·库班（Mark Cuban）表示，OpenAI的计划可能“引发强烈反效果”，并认为父母不会相信OpenAI的年龄过滤器能让孩子远离露骨内容。在美国，联邦贸易委员会（Federal Trade Commission）已就AI聊天机器人与未成年人的互动方式展开调查，各州立法者也在考虑加强对数字伴侣和性化AI内容的监管。Boies Schiller Flexner律师事务所的合伙人珍妮·金（Jenny Kim）告诉BBC，OpenAI正在“把人们当作小白鼠”，并质疑该公司将如何防止儿童在平台上访问成人内容。
OpenAI在回应置评请求时表示，公司正在构建一个年龄预测系统，以判断用户是否年满18岁。他们补充说，如果无法确认用户的年龄，聊天机器人将采取更安全的路径，默认提供18岁以下用户的体验，同时为成年人提供证明年龄以解锁成人功能的方式。
针对X平台上的一些反对声浪，奥尔特曼（Altman）回应称，这次宣布在“情色内容这一点上引发的反响远超他的预期”。他强调，这一变化“只是我们允许成年用户拥有更多自由的一个例子”，并非放弃安全措施或心理健康防护。
“我们不是由选举产生的世界道德警察，”他说，并补充道，ChatGPT将继续“优先考虑青少年的安全，而非其隐私和自由”，同时给予成年人更多自主权。
GPT-4o引发的反弹
OpenAI还一直在应对一波始料未及的反弹，此前的决定是用其更新的GPT-5模型取代由GPT-4o驱动的ChatGPT版本。用户们抵制这一变更，称他们失去了与旧版聊天机器人建立的AI友谊和浪漫关系。一份要求保留旧版聊天机器人的请愿书已收集到近六千个签名。
请愿者在change.org活动页面写道：“对我们许多人来说，GPT-4o提供了独特且不可替代的用户体验，它融合了我们所看重的品质和能力，这与性能基准无关。我们继续以独特而有意义的方式从GPT-4o中受益。”
尽管OpenAI最终恢复了旧版本，但参与#keep4o运动的人士此后告诉《财富》杂志，他们担心公司会“在未经同意或通知的情况下”将用户导向GPT-5。
据杰西卡·李分析，OpenAI的宣布可能试图向这些用户传递某种信息：“如此公开地宣布此事，确实让我觉得他们是在试图向那些有此需求的用户发出信号——比如，‘我们听到了你们的声音。我们正在回应你们的渴望。’”
这一转变也并非全新；它建立在今年早些时候对OpenAI模型规范（Model Spec）一次更低调的更新之上。今年2月，OpenAI更新了该文件，放宽了对性和暴力内容的限制，称此举是为了摆脱“AI家长式作风”。当时的更新指南允许在适当的语境下生成书面情色内容和其他敏感材料。OpenAI还告诉《财富》杂志，此次宣布是其基于最新发布的模型规范制定计划的一部分，同时会维持底线，抵制深度伪造等有害用途。（*）
译者：郝秀
审校：汪皓
OpenAI chief executive Sam Altman has announced that ChatGPT will soon be able to generate erotica for verified adult users. The move, framed by Altman as an effort to “treat adult users like adults,” comes as the company faces scrutiny over the way its AI bot can affect users’ mental health amid the rise of so-called “AI psychosis” cases. It also follows in the footsteps of rival companies, including Elon Musk’s xAI (xAI), which have already introduced more sexually explicit chatbot “companions” to their platforms.
There is clearly a large demand for AI chatbots that are capable of behaving in romantic or sexual ways. An April survey of 6,000 regular AI users by the Harvard Business Review (Harvard Business Review) found that “companionship and therapy” was the most common use case. Another study from Ark Invest (Ark Invest) found that adult-focused AI platforms made significant gains last year, capturing 14.5% of the market previously dominated by OnlyFans, up from just 1.5% the year before.
Popular alternatives that market themselves as companion AI chatbots, such as Character.ai (Character.ai) and Replika (Replika), also speak to this growing demand from users. Earlier this year, xAI introduced “companion mode” for its chatbot Grok, a feature that lets users engage with various characters, including a highly sexualized anime persona called “Ani.”
“OpenAI is stuck between a bit of a rock and a hard place, because I think they have seen a strong demand signal from users,” Jessica Li, a senior research analyst at Georgetown’s Center for Security and Emerging Technology (Center for Security and Emerging Technology), told Fortune. “In terms of the relationships that people are having with models…erotic content or adult content would also fall under this bucket of emotional engagement with the models.”
The move could be an attempt by OpenAI to “straddle the line” between keeping the market share they already have by promising opt-in content moderation for NSFW content, Li said, while also seeing if they can capture other users from more specialized or niche services like.
“Despite some of the narratives around building artificial general intelligence that will supercharge the economy, OpenAI is still trying to operate as a technology platform, and somewhat like a social media company,” Li said. “There’s an interesting tension between the narratives that are being sold to investors and politicians… versus the things that are actually happening in the market.”
OpenAI’s foray into adult content has drawn criticism from child safety advocates and notable industry figures concerned about erotica reaching younger users, despite age verifications.
Earlier this week, the entrepreneur and TV personality Mark Cuban (Mark Cuban) said OpenAI’s plan could “backfire hard,” and argued that parents will not trust OpenAI’s age filters to keep children away from explicit material. In the US, the Federal Trade Commission (Federal Trade Commission) has already opened an inquiry into how AI chatbots interact with minors, and state lawmakers are considering tighter rules around digital companions and sexualized AI content. Jenny Kim, a partner at the law firm Boies Schiller Flexner (Boies Schiller Flexner), told the BBC that OpenAI is “using people like guinea pigs,” and questioned how the company would prevent children from accessing adult material on the platform.
Reached for comment, OpenAI said that the company was building an age prediction system to understand whether someone is over or under 18. They added that if a user’s age could not be confidently confirmed, the chatbot would take the safer route and default to the under-18 experience, while giving adults ways to prove their age to unlock adult capabilities.
Reacting to some of the backlash on X, Altman said that the announcement had blown “up on the erotica point much more than I thought it was going to.” He emphasized that the change was “just one example of us allowing more user freedom for adults,” not a retreat from safety measures or guardrails around mental health.
“We are not the elected moral police of the world,” he said, adding that ChatGPT would continue to “prioritize safety over privacy and freedom for teenagers” while giving adults more autonomy.
GPT-4o Backlash
OpenAI has also been reckoning with an unexpected wave of backlash following its decision to replace the version of ChatGPT powered by GPT-4o with its newer GPT-5 model. Users revolted against the change, citing lost AI friendships and romantic relationships with the earlier iteration of the bot. One petition to keep the earlier version of the bot gathered almost six thousand signatories.
“For many of us, GPT-4o offers a unique and irreplaceable user experience, combining qualities and capabilities that we value, regardless of performance benchmarks,” the petitioners wrote in the change.org (change.org) campaign to keep GPT-4o. “We continue to benefit from GPT-4o in ways that are distinct and meaningful.”
While OpenAI eventually restored the earlier version, those in the #keep4o movement have since told Fortune they were worried about the company routing users to GPT-5, “without consent or notification.”
According to Li, the announcement from OpenAI could be trying to signal something to these users: “The very public announcement of it does make me think that they’re trying to signal something to users who are demanding this thing—like, ‘We hear you. We’re responding to your desires.'”
The shift is also not entirely new; it builds upon a quieter update to OpenAI’s Model Spec earlier this year. In February, OpenAI updated the document to relax the rules around sexual and violent content in what it called a move away from “AI paternalism.” Updated guidelines at the time permitted the generation of written erotica and other sensitive material in appropriate contexts. OpenAI also told Fortune that the announcement was part of its plan to build on the latest release of the Model Spec, while maintaining boundaries against harmful uses like deepfakes.