广告
加载中

别想用软色情做智能体的跳板

陈玉立 2026-07-06 10:33
陈玉立 2026/07/06 10:33

邦小白快读

EN
全文速览

本文核心是7月15日《人工智能拟人化互动服务管理暂行办法》正式施行前,豆包、千问两大厂先后下线主App中的关系型智能体功能,并非智能体赛道整体退潮,只是软色情、擦边灰色增长部分退潮,给普通用户的核心干货如下:

1. 智能体行业接下来会分层发展,办公协作、写代码、做纪要这类任务型效率智能体不会受影响,会继续正常发展,普通用户可以放心使用这类提升效率的产品。

2. 情感陪伴、虚拟恋人这类关系型智能体不会消失,但会进入强监管阶段,软色情、擦边、侵权仿冒的内容会被大规模清理,用户要自觉远离违规内容,未成年人需要家长做好引导防护。

3. 结合AI情感陪伴的情欲交互硬件已经被新规明确禁止,后续很难上市流通,普通用户不要购买这类违规产品,避免自身权益受损。

新规落地后智能体赛道迎来重新洗牌,给AI相关品牌商的干货总结如下:

1. 消费趋势方面,孤独经济带来的情感陪伴需求确实真实存在,但依靠软色情、擦边内容拉动增长的路径已经完全走不通,品牌不能再用灰色内容做智能体业务的增长跳板。

2. 产品研发层面,布局关系型智能体业务必须把合规成本纳入预算,提前补上年龄分级、未成年人保护、内容审核、角色版权审核、诱导消费管控等合规模块,满足监管要求。

3. 品牌风险层面,做大众AI工具的主品牌,要将高风险的情感互动功能剥离出主App,避免违规内容风险损害主品牌的公共信任,影响整体业务发展。

4. AI硬件品牌需要提前明确产品定义,不能借陪护、情绪支持的名义包装情欲互动产品,否则会直接面临合规障碍,产品无法上市。

本次新规落地带来智能体赛道的大调整,给相关AI业务卖家的干货总结如下:

1. 政策解读:本次新规没有停留在内容审核层面,而是把监管边界延伸到了产品形态,明确规定情感陪护可以合规开展,但借陪护关系导向情欲消费的行为被禁止,同时明确禁止提供亲密性模拟交互内容或硬件功能。

2. 机会提示:办公协作、企业服务、效率工具类的任务型智能体不受本次监管冲击,仍然属于有潜力的增长市场,卖家可以重点布局这类业务。

3. 风险提示:过去关系型智能体野放式增长的模式已经走不通,合规成本大幅上升,没有能力承担全链路审核、风控成本的小团队,建议谨慎入场。

4. 做AI硬件的卖家需要重新评估产品定位,带亲密模拟结构的产品已经不符合合规要求,要及时调整方向,避免开模生产后产生巨额损失。

新规实施后智能体赛道出现明确分流,给AI相关生产工厂的干货总结和启示如下:

1. 产品生产设计需求已经发生变化,原本计划推进的“拟人化AI+情欲硬件”项目已经触碰监管红线,不能继续开模生产,需要立刻调整设计方向,剥离亲密交互相关的结构和功能。

2. 商业机会方面,任务型智能体配套硬件、合规运营的情感陪护硬件,比如正规老人照护终端、儿童辅助设备仍然有稳定的市场空间,工厂可以主动对接合规品牌方获取订单。

3. 做AI相关业务的启示:工厂布局AI相关产品,必须在设计阶段就明确产品定位,提前开展合规风险评估,不能跟风布局灰色项目,否则产品生产铺货后整改成本极高,会造成难以挽回的损失。

4. 工厂需要建立产品设计阶段的合规审核机制,提前匹配监管要求,降低产品上市后的合规风险。

新规落地后智能体行业正式进入合规化发展阶段,给AI相关服务商的干货总结如下:

1. 行业发展趋势:智能体赛道不会整体降温,只会进入分层发展阶段,任务型智能体加速发展,关系型智能体进入强合规时代,灰色的拟人AI+情欲硬件会被行业出清,合规已经成为全行业的核心刚性需求。

2. 客户核心痛点:当前AI平台和品牌方的核心痛点转变为如何管控合规风险,关系型智能体平台需要覆盖内容审核、版权审核、未成年人保护、防沉迷、风险干预等全链路合规能力,大部分中小玩家没有能力独立搭建这套体系。

3. 新的商业机会:服务商可以针对性开发合规相关解决方案,比如给关系型智能体平台提供违规内容审核服务、用户身份年龄验证服务、角色版权核验工具,给AI硬件厂商提供产品合规评估咨询服务,这些都是新增的市场需求,有较大的增长空间。

此次豆包、千问调整智能体业务的动作,给AI平台商的干货提示如下:

1. 风险规避方面,做大众AI助手的主平台,要将高风险的关系型智能体、UGC智能体广场、开放情感互动功能从主入口剥离,不能再把高风险互动当成低成本增长工具,避免合规风险牵连主平台,损害平台的公共信任。

2. 运营管理调整:如果平台选择保留关系型智能体业务,必须尽快完善运营规则,把年龄分级、内容审核、版权审核、防沉迷、诱导消费管控、风险提示等合规要求融入全运营流程,落实平台的分发管理责任。

3. 招商调整:平台不能再引入主打擦边亲密互动的关系型智能体项目,也不能引入不合规的AI情欲硬件商家,避免平台因为分发违规内容承担监管责任。

4. 业务方向调整:平台可以重点倾斜资源发展任务型智能体业务,这类业务风险清晰、需求稳定,符合监管方向,不会带来额外的合规压力。

此次新规落地和头部大厂的业务调整,反映了AI智能体产业的最新发展动向,给产业研究者的干货总结如下:

1. 产业新动向:智能体产业正式告别野蛮生长,进入合规化分层发展阶段,产业不会整体降温,任务型智能体会继续加速发展,关系型智能体进入强合规运营阶段,灰色情欲相关智能产品会被行业出清,合规能力、风控体系已经成为新的核心竞争力。

2. 新出现的产业问题:关系型智能体原来的商业模式建立在情绪依赖和低成本UGC增长基础上,合规成本显性化后,原有商业模式需要重构,中小团队的轻资产快速增长模式难以为继,行业集中度会提升。

3. 政策启示:本次监管没有停留在内容治理层面,而是延伸到产品定义和产品形态,给AI行业监管提供了新的范式,也给产业研究提出了新议题,需要研究合规成本对AI创新的影响,以及如何平衡真实需求满足和社会风险管控。

4. 后续可以重点关注合规框架下关系型智能体的可持续商业模式,以及任务型智能体的产业落地路径等方向。

返回默认

声明:快读内容全程由AI生成,请注意甄别信息。如您发现问题,请发送邮件至 run@ebrun.com 。

我是 品牌商 卖家 工厂 服务商 平台商 研究者 帮我再读一遍。

Quick Summary

Ahead of the July 15 implementation of China's Interim Measures for the Administration of Artificial Intelligence Anthropomorphic Interactive Services, Doubao and Qianwen, two leading Chinese AI agents operated by major tech firms, have successively removed relational AI agent features from their main mobile apps. This does not signal a broad downturn for the entire AI agent track; rather, it marks a contraction in the grey-market growth segment centered on soft pornography and suggestive content. Key takeaways for general users are as follows:

1. The AI agent industry will develop along stratified lines going forward. Task-oriented efficiency agents for office collaboration, code writing, meeting minute compilation and other similar use cases will face no disruption and continue normal development. General users can safely use these products to boost productivity.

2. Relational AI agents such as those for emotional companionship and virtual dating will not disappear, but will enter a period of strict regulation. Soft pornography, suggestive content, infringing and counterfeit agents will be cleared out on a large scale. Users should consciously stay away from non-compliant content, and parents need to provide proper guidance and protection for minors.

3. Erotic interactive hardware integrated with AI emotional companionship is explicitly banned under the new regulation, and will be very difficult to bring to market or circulate going forward. General users should not purchase these non-compliant products to avoid risking harm to their own rights and interests.

The implementation of the new regulation has triggered a reshuffle of the AI agent track. Key takeaways for AI-related brands are as follows:

1. In terms of consumer trends, while demand for emotional companionship driven by the 'loneliness economy' is genuine, growth strategies relying on soft pornography and suggestive content are no longer viable. Brands can no longer use grey-market content as a growth springboard for their AI agent businesses.

2. For product R&D, brands developing relational AI agent businesses must incorporate compliance costs into their budgets, and pre-build compliance modules including age rating, minor protection, content review, character copyright review and induced consumption control to meet regulatory requirements.

3. In terms of brand risk, main brands offering mass-market AI tools should spin off high-risk emotional interaction features from their main apps, to avoid damage to the core brand's public trust from non-compliant content that could harm overall business operations.

4. AI hardware brands must clarify their product definitions early on. They must not package erotic interactive products under the guise of companionship or emotional support, as this will directly lead to compliance barriers and prevent products from launching.

The rollout of the new regulation has triggered major adjustments to the AI agent track. Key takeaways for sellers of related AI businesses are as follows:

1. Policy interpretation: The new regulation does not stop at content review; it extends regulatory boundaries to product forms. It explicitly allows compliant emotional companionship services, but bans the practice of driving erotic consumption under the guise of companionship relationships. It also explicitly prohibits sexually-simulated interactive content or hardware functions.

2. Opportunity note: Task-oriented AI agents for office collaboration, enterprise services and productivity tools face no impact from this round of regulation, and remain a high-potential growth market. Sellers should prioritize布局 these business lines.

3. Risk note: The previous unregulated growth model for relational AI agents is no longer viable, and compliance costs have risen sharply. Small teams unable to bear the cost of full-link review and risk control are advised to enter this space with caution.

4. AI hardware sellers need to re-assess their product positioning. Products with sexually-simulating structures no longer meet compliance requirements, and sellers should adjust direction promptly to avoid massive losses after completing mold production.

The implementation of the new regulation has driven a clear divergence in the AI agent track. Key takeaways and insights for AI-related manufacturing factories are as follows:

1. Product design and manufacturing demand has shifted. Originally planned 'anthropomorphic AI + erotic hardware' projects have crossed the regulatory red line, and cannot proceed with mold opening and production. Factories must adjust their design directions immediately, and remove structures and functions related to intimate interaction.

2. In terms of business opportunities, supporting hardware for task-oriented AI agents and compliant emotional companionship hardware—such as formal elder care terminals and children's assistive devices—still maintain stable market space. Factories can actively partner with compliant brand owners to secure orders.

3. Key insight for AI-related business: When developing AI-related products, factories must clarify product positioning at the design stage, conduct compliance risk assessments in advance, and avoid jumping on the bandwagon for grey-market projects. Otherwise, rectification costs after production and distribution will be extremely high, leading to irreversible losses.

4. Factories need to establish a compliance review mechanism at the product design stage, align with regulatory requirements in advance, and reduce compliance risks after products launch.

With the rollout of the new regulation, the AI agent industry has officially entered a stage of compliant development. Key takeaways for AI-related service providers are as follows:

1. Industry development trend: The AI agent track will not cool across the board, but will enter a phase of stratified development: task-oriented agents will accelerate growth, relational agents will enter an era of strict compliance, and grey-market anthropomorphic AI + erotic hardware will be cleared from the industry. Compliance has become a core, non-negotiable demand across the sector.

2. Core customer pain points: The core priority for AI platforms and brands has shifted to managing compliance risk. Relational AI agent platforms need full-link compliance capabilities covering content review, copyright review, minor protection, anti-addiction, and risk intervention. Most small and mid-sized players lack the capacity to build this system independently.

3. New business opportunities: Service providers can develop targeted compliance solutions. This includes illegal content review services, user identity and age verification services, and character copyright verification tools for relational AI agent platforms, as well as product compliance assessment and consulting services for AI hardware manufacturers. All of these are newly emerging market demands with significant growth potential.

The adjustments to AI agent businesses by Doubao and Qianwen offer the following key insights for AI platform operators:

1. Risk mitigation: Main platforms serving mass-market AI assistants should strip high-risk relational agents, UGC agent marketplaces, and open emotional interaction features from their main entry points. They can no longer treat high-risk interaction as a low-cost growth tool, to avoid compliance risks dragging down the main platform and damaging its public trust.

2. Operational adjustments: If a platform chooses to retain relational AI agent business, it must quickly improve operational rules, integrate compliance requirements including age rating, content review, copyright review, anti-addiction, induced consumption control and risk warnings into full operational processes, and fulfill the platform's responsibility for distribution management.

3.招商 adjustments: Platforms must no longer onboarding relational AI agent projects focused on suggestive intimate interaction, nor non-compliant AI erotic hardware sellers, to avoid the platform being held regulatory responsible for distributing non-compliant content.

4. Business direction adjustment: Platforms can reallocate resources to prioritize developing task-oriented AI agent businesses. These businesses have clear risk profiles, stable demand, align with regulatory guidance, and do not bring additional compliance pressure.

The implementation of the new regulation and business adjustments by leading industry players reflect the latest development trends of the AI agent industry. Key takeaways for industry researchers are as follows:

1. New industry trends: The AI agent industry has officially left behind its era of unregulated growth and entered a phase of compliant, stratified development. The industry will not cool overall: task-oriented agents will continue to accelerate growth, relational agents will enter a phase of strictly compliant operations, and grey erotic-related intelligent products will be cleared from the industry. Compliance capabilities and risk control systems have become the new core competitiveness.

2. New emerging industry issues: The original business model for relational AI agents was built on emotional dependency and low-cost UGC-led growth. After compliance costs have become explicit, the original business model requires restructuring. The light-asset, rapid growth model used by small and mid-sized teams is no longer viable, and industry concentration will increase.

3. Policy insights: This round of regulation does not stop at content governance, but extends to product definition and product form, creating a new paradigm for AI industry regulation and raising new research topics for industry scholars. Research is needed on the impact of compliance costs on AI innovation, and how to balance the satisfaction of genuine consumer demand with social risk management.

4. Future priority research areas include the sustainable business model for relational AI agents under a compliance framework, and the industry landing path for task-oriented AI agents.

Disclaimer: The "Quick Summary" content is entirely generated by AI. Please exercise discretion when interpreting the information. For issues or corrections, please email run@ebrun.com .

I am a Brand Seller Factory Service Provider Marketplace Seller Researcher Read it again.

出品|虎嗅ESG组

作者|陈玉立

本文是#ESG进步观察#系列第169篇文章

本次观察关键词:供应链治理

近日,豆包与千问几乎同时宣布下线智能体功能。

据公开报道,豆包智能体下线后,相关功能存在于独立产品“猫箱”;而千问处理得更彻底,直接关闭相关功能且不提供迁移路径。

两家大厂不仅处置方式类似,时间点也高度一致——都卡在7月15日《人工智能拟人化互动服务管理暂行办法》正式施行之前,办法提到需对虚拟恋人、深度角色扮演、情感依附类AI设立强约束。

因此,大厂切掉的并非问答、办公、搜索、代码、工作流这类通用AI能力,真正被重新定价的是过去一两年里快速膨胀的关系型智能体:用户自建人设、角色扮演、虚拟恋人、情感陪伴、公开分发的智能体广场。

一句话总结:智能体没有整体退潮,退潮的是低门槛UGC、情绪依赖和灰色亲密互动撑起来的那部分增长。

在新规之后,智能体赛道可能开始分流:任务型智能体继续发展,关系型智能体进入强合规时代,带情感陪伴功能的AI“情趣”硬件大概率被堵在门外。

大厂切割“关系型智能体”

在豆包与千问的动作中,值得观察的地方在于“下线的究竟是什么”。

从平台处理方式看,豆包并没有放弃大模型主产品,也没有放弃AI助手能力,反而选择把高风险互动功能从主App中剥离,只在独立产品中存续。这一动作的含义明确:主App要继续承担大众效率工具、搜索入口、AI助手的角色,情感互动、角色扮演、用户生成智能体这类风险更高的功能,则被放到“隔离区”。

千问的处理则是直接关闭,且不给迁移路径,说明平台对这块业务的判断更冷。如果一个非核心功能带来的审核成本、监管压力、舆论风险超过业务收益,关掉就是最干净的选择。

此次下架动作中两家底层逻辑一致:关系型智能体在大厂主入口里的位置变了。过去它能带来互动、时长和用户参与;现在,它开始变成风险资产。

一个值得追问的议题是,为什么关停智能体的是豆包、千问?为什么WorkBuddy这类产品看起来没有受到同样冲击?任务型Agent,还有谁在做?

答案大概率藏在产品属性里。

WorkBuddy这类偏办公、企业协作、工作流的智能体,处理的是任务:写文档、做纪要、调流程、跑数据。它的主要风险集中在数据权限、企业信息安全、输出准确性和流程责任上;而关系型智能体处理的是人和拟人对象之间的关系,用户完全可以创建一个“AI女友”“AI男友”“病娇妹妹”“霸总伴侣”,加上平台再提供广场、推荐、和分享,这就从工具变成了拟人化内容平台,风险也就随之放大。

首先可能会带来UGC失控。用户可以低成本创建人设、设定语气、关系、头像、开场白,再公开发布。供给越丰富,审核压力越大。此前上海累计下架1.4万个违规智能体,就是一个很直接的信号:不加管制的开放智能体生态容易滋生擦边内容、侵权人设、真人仿冒和低俗互动。

其次是关系依赖。普通问答助手回答完就结束,关系型智能体会持续记忆、主动关心、使用亲密称呼、模拟陪伴关系。用户投入越深,产品留存越强,付费转化也越容易。但在未成年人、心理脆弱人群、孤独老人等场景里,这种黏性很容易越过安全边界。

最后是平台分发责任。一个用户自己创建、自己使用的智能体,风险半径有限;一旦进入智能体广场、排行榜、搜索页和推荐流,平台就参与了放大。此时再用“用户自建”解释责任,就有些“无力”了。

这就是豆包、千问这类大流量入口最敏感的地方,它们做全民AI助手的品牌要求是干净、可信、普适。但主App里如果混入大量虚拟恋人、暧昧陪伴、侵权角色、擦边人设,一旦被截图传播,损害的会是整个产品的公共信任。

豆包和千问的下架动作本质上都是停止为非核心场景承担额外风险。两者都说明,大厂已经开始给关系型智能体重新算账。

新规把边界画到了产品形态上

新规真正有杀伤力的地方,在于它没有停留在“内容审核”层面,而是把监管边界延伸到了服务定义和产品形态。

据《人工智能拟人化互动服务管理暂行办法》第八条指出,生成、提供淫秽色情、亲密性模拟交互内容或硬件功能,被明确禁止。这条规则直击“拟人化关系商业化”的灰色地带。

过去很多AI陪伴产品会用“情感”“陪护”“孤独经济”包装自己。表面上它们提供倾诉、陪伴、安慰,实际上增长最强的是更暧昧的部分:不限于亲密称呼、恋爱角色、暧昧对话等。用户越投入,停留越久;关系越强,付费越容易。

关系型智能体把AI从工具变成关系对象,把一次性的使用变成持续性的陪伴。会员、角色权益、高级模型、道具、订阅,都可以接在这条关系链上。

问题也出在这里。

当产品开始利用孤独、亲密需求、情绪脆弱性提高留存和付费,监管关注点就会从“模型有没有说错话”转向“平台有没有制造依赖”。如果再叠加未成年人、诱导消费、色情暗示、侵权人设、心理疏导越界,风险等级会迅速上升。

所以,新规划出的边界可以概括成一句话:情感陪护可以做,把陪护关系性化不行;拟人化交互可以做,用拟人化关系导向情欲消费不行。

从这一角度看,对AI硬件公司的压力更大。

很多硬件从业者原本以为,拟人化互动监管主要影响聊天App、智能体平台、虚拟陪伴软件。但新规覆盖的是“拟人化互动服务”,硬件只要搭载语音陪伴、主动关怀、仿人外观,就可能进入监管视野。

尤其是AI陪护机器人、AI宠物、AI玩具、老人陪伴终端等。这些产品天然会强调陪伴感。为了提高用户黏性,厂商会设计专属称呼、情绪识别、主动问候、记忆偏好、拟人语气甚至肢体动作。这些设计本身并不必然违规,但边界会变得非常敏感。

儿童场景涉及未成年人保护、防沉迷、价值观引导、诱导消费等;老人场景涉及认知能力下降、健康建议误导、诈骗风险等。只要厂商宣传自己提供“陪护”“心理疏导”“情绪安抚”,就不能再把自己简单定义成卖硬件。

更高风险的是“情感陪伴AI+性交互硬件”。

成人硬件行业过去有一个明显痛点:一次性消费多,持续留存弱,用户关系浅。AI给它提供了新的想象空间:让硬件拥有人格、记忆和情绪回应,把工具变成“伴侣”,再通过订阅、内容、互动脚本实现持续变现。

新规直接堵住了这条路。

一款产品如果同时具备AI情感陪伴、长期人格记忆、情欲互动算法、仿真性结构或模拟性行为机械结构,就很难再被解释成普通成人用品或普通陪护设备。它的风险不再是模型偶尔越界,而是产品设计目标本身踩线。新规施行后,这类产品在安全评估、上市销售、渠道合规上都会遭遇实质障碍。

软件产品出问题,还能下线入口、调整模型、增加过滤、修改提示词。硬件一旦开模、生产、铺货、上市,整改成本极高。产品形态如果已经带着亲密模拟结构,合规风险就被固化在设备上。

所以AI硬件从业者的压力,来自一个更前置的问题:产品定义能不能经得起审查。

你到底是生活辅助设备,还是虚拟伴侣?是老人照护终端,还是借陪护之名包装亲密互动?是情绪支持工具,还是通过人格化关系推动情欲消费?

这些问题回答不清,产品很难走远。

智能体开始分流

一个判断是,智能体行业接下来不会整体降温,但会明显分层。

第一类,任务型智能体会继续加速。企业Agent怎么控风险?

办公助手、企业知识库、代码智能体、客服助手、会议纪要、数据分析助手、流程自动化Agent,这些产品解决的是明确任务。它们的价值可以用效率、成本、准确率、响应速度、流程节省来衡量。

这类产品的合规重点主要在数据安全、权限控制、企业审计、输出责任。它们不会主动制造亲密关系,也不依赖用户情绪黏性变现。WorkBuddy这类产品能相对平稳,原因就在这里。

它有风险,但风险类型清楚;它要治理的是“组织如何安全使用AI”,不是“用户是否对一个虚拟人格产生依赖”。

第二类,关系型智能体进入强合规时代。

AI恋人、虚拟男友/女友、情绪陪伴、角色扮演、用户自建智能体、智能体广场,并不会因为新规直接消失。陪伴、倾诉、孤独缓解、养老辅助、心理支持,都是真实需求。AI在这些场景里也有价值。

但过去那套玩法走不通了。

开放UGC、擦边人设、低成本角色库、热门榜分发、亲密互动拉留存、会员订阅收割情绪依赖,这条链路会越来越难做。平台必须补上年龄分级、防沉迷、未成年人保护、内容审核、真人仿冒治理、角色版权审核、风险提示、诱导消费限制、情感依赖边界管理等。

而这些东西也都会进入成本表。

过去一个小团队接入模型API,做一套角色模板,加上广场和分享机制,很快就能跑出一个陪伴产品。合规成本没有显性化时,增长故事很好讲。一旦审核、投诉、版权、未成年人保护、心理风险干预都变成硬成本,很多产品的商业模型会被重新计算。

关系型智能体仍有机会,但会从“野放式增长”进入“强边界运营”。留下来的,不会是最会撩拨情绪的团队,而是能把场景、责任和风险讲清楚的团队。

第三类,“拟人化AI+情欲硬件”会被挤出主流市场。

这类产品的问题最尖锐。聊天产品偶发越界还能补救,硬件产品一旦把仿真结构、亲密模拟、情欲算法和AI人格绑定起来,就已经完成了高风险产品闭环。

新规拆掉了很多灰色包装:你不能一边宣称养老陪护、心理疏导、日常陪伴,一边设计情欲互动算法和性交互硬件;不能一边享受拟人化关系带来的高黏性,一边又把风险藏在“成人自愿使用”的叙事里。

这显然会带来一轮行业出清。短期内,平台会收缩智能体业务,清理敏感角色,降低情感陪伴功能在主App里的权重;AI陪伴产品会强化分级、审核和提示;部分AI硬件项目会推迟上市,重新评估外观、交互和功能边界。

长期看,智能体行业会回到更清晰的价值框架里:能提高效率的继续做,能提供照护的规范做,依赖擦边和情欲变现的退出。

这也是豆包、千问这次动作给行业的最大信号。

它们并没有否定AI应用,也没有否定智能体方向。它们只是用产品调整告诉行业:关系型智能体不能再无差别地塞进主入口,平台不能继续把高风险互动当成低成本增长工具。

AI行业进入下一阶段后,竞争不只看模型能力、入口流量和交互体验,还要看合规能力、风控体系和社会信任。

过去,智能体被想象成一个“什么都能扮演、什么关系都能模拟”的万能入口。新规之后,这一想象需要重写。所谓“智能体告别野蛮生长”,说到底,是行业开始为产品边界、用户脆弱性和商业模式付账。

注:文/陈玉立,文章来源:虎嗅APP(公众号ID:huxiu_com),本文为作者独立观点,不代表亿邦动力立场。

文章来源:虎嗅APP

广告
微信
朋友圈

FAQ回顾

关系型智能体是什么?

关系型智能体是具备用户自建人设、角色扮演、虚拟恋人、情感陪伴、公开分发功能的AI产品,区别于处理办公、搜索、代码等明确任务的任务型智能体,此前主要依赖低门槛UGC、情绪依赖、灰色亲密互动实现快速增长。

人工智能拟人化互动新规施行后智能体赛道有哪些变化?

新规施行后智能体赛道将出现明显分流:任务型智能体将持续加速发展,合规重点集中在数据安全、权限控制等领域;关系型智能体进入强合规时代,需补足年龄分级、防沉迷等风控能力;带有情欲互动功能的AI硬件将被挤出主流市场。

为什么豆包、千问要下线相关智能体功能?

豆包、千问的下线动作均发生在《人工智能拟人化互动服务管理暂行办法》正式施行前,关系型智能体带来的审核成本、监管压力、舆论风险已超过业务收益,平台选择停止为非核心场景承担额外风险,保障主产品的公共信任。

这么好看,分享一下?

朋友圈 分享

APP内打开

赞 +1
+1
微信好友 朋友圈 新浪微博 QQ空间
关闭
收藏成功
发送
/140 0