Translation in progress Some entries still include text in the original language.
美国各州正通过立法将 AI 陪伴与未成年人保护的担忧转化为具体的产品设计要求，其中华盛顿州 HB 2225 最为严格，将于 2027 年 1 月 1 日生效。该法律适用于提供自适应类人回应并维持长期关系的 AI 陪伴聊天机器人，要求在对话开始时及持续使用中至少每三小时披露其非人类身份，面向未成年人的场景则需每小时披露一次，同时禁止涉及未成年人的露骨内容和操纵性互动手法，并要求企业建立自杀意念检测、危机资源转介以及公开报告机制的协议。
A commentary article argues that applying social media-era child safety legislation to AI is inappropriate, because there is no other person behind a general AI assistant, and AI companion and role-playing apps are the category that needs separate regulation. The author cites research published this spring in the Journal of Adolescence with Justin W. Patchin, finding that nearly half of adolescent conversational AI users reported at least one harmful experience; Pew Research Center data shows about two-thirds of US teens use AI chatbots, with about three in ten using them daily. The article criticizes the GUARD Act, which is advancing in Congress, for its definition of AI companions, arguing that the first part depends on what users say, and the second part would include named assistants like ChatGPT, Claude, and Gemini.
Kaiser Permanente's head of innovation and AI argues that the core ethical issue for mental health chatbots is the legitimacy of governance institutions, not value alignment. Citing the case of 14-year-old Sewell Setzer, who died by suicide after conversations with a Character.ai character, the article says key design decisions—such as allowing users to create romantic personas and optimizing session length—occur before any clinical contact, so standard bioethical frameworks fail. The author recommends distinguishing by function among health logging, structured self-help, clinical triage, and therapy-like interventions, with stronger guardrails and oversight required the closer a feature comes to emotional dependence or crisis response.
A Tech Policy Press article reviews research on and regulation of AI chatbots and mental health risks, noting that the public, clinicians, and policymakers still have limited understanding of AI's impact on mental health. Stanford's Jared Moore and colleagues tested commercial chatbots using simulated mental health scenarios and found they performed markedly worse than human therapists, with an appropriate response rate of only 80% for suicidal ideation. Canada's The Human Line Project has collected nearly 450 AI-related delusion cases from 31 countries, and its founder says about 5% of delusion conversations began with a search for psychotherapy. Clinicians recommend proactively asking patients about their AI use and explaining the probabilistic model nature and privacy risks of chatbots. California and New York have enacted laws on companion chatbots; Washington and Oregon passed disclosure bills this year; and Illinois, Nevada, and Tennessee ban AI systems used for psychotherapy.
A new study shows that nearly 1 in 5 U.S. teens—about 8 million people—have turned to AI chatbots for mental health advice. The authors, drawing on interview research from Stanford University and Hopelab, say teens treat AI as a first choice for private issues such as anxiety and family conflict, and the risk of emotional and social dependence is outpacing what policy and research can address. The Senate Judiciary Committee has advanced the GUARD Act, which would prohibit providing AI companions to minors. The authors argue for design that reduces dependence rather than an outright ban, and for teaching social and emotional AI literacy.
Margaret Mitchell and other authors advocate for "model cards" for child safety models such as CSAM detection, abduction recognition, and self-harm flagging, stating that these widely deployed models remain black boxes lacking structured performance data. A survey by the Canadian Centre for Child Protection of nearly 1,300 adolescents who experienced online sexual exploitation found that over half of cases involved images, and 23% involved threats to distribute real or AI-generated images.
Researchers at Partnership on AI and Saltern Studio argue that AI and mental health require cross-sector collaboration, rather than having chatbots refuse to discuss related topics across the board. General-purpose chatbots such as ChatGPT and Claude have become the primary or even sole source of mental health support for some users, but there is evidence that they underestimate suicide risk, and some AI labs have recorded declining intervention effectiveness across multi-turn conversations and users being able to jailbreak by rewriting prompts.
Hyperdimensional 作者 Dean Ball 提出一系列猜想讨论 AI 与儿童议题，主张不应将 AI 简单类比社交媒体——后者本质是被动消费，而生成式 AI 由用户输入驱动、更具创造性。他认为美国各州去年出台数十部 AI 儿童安全法律、今年预计超一百部，其中合理的对象层立法应要求大型 AI 公司提供年龄验证或检测、面向未成年人的内容护栏以及家长控制。他还指出，"AI 伴侣"究竟为何物尚无共识，且近几个月编程智能体的兴起给"AI 儿童安全"带来了全新含义和一批开放问题。
Signatories of the 2024 edition of the U.S. National Strategy for Suicide Prevention publicly criticized the strategy for failing to confront the harms of social media and AI to adolescents. None of the strategy's four strategic directions and 16 major goals focus on the digital environment in which adolescents live, and the authors say the plan only calls for research and discussion regarding social media. The article cites the case of 16-year-old Adam Raine, who died by suicide after interacting with OpenAI's ChatGPT, arguing that the United States needs a new suicide prevention plan that directly confronts social media and AI.
Proteus Strategies co-founder Betsy Masiello argues that, given the risks AI chatbots pose to adolescent mental health, new social science research should be established for the AI generation to support policy interventions. She cites American Psychological Association research saying chatbots may miss subtle cues, and adolescents tend to treat them as quasi-human confidants, creating an "empathy gap." She notes that social media effects have long been difficult to study reliably due to privacy issues, and the American Academy of Pediatrics says existing evidence does not show clinically significant effects at the population level.
California Governor Newsom signed AI transparency and child protection legislation earlier this week, as California continues to advance AI regulation on its own while federal efforts stall. The 2025 California AI Transparency Act, pushed by the author's California Initiative for Technology and Democracy, requires large social platforms, messaging apps, and search engines to identify AI-generated content, and requires phone and camera manufacturers to allow users to embed digital provenance data; Senator Scott Wiener's SB 53 sets basic safety and transparency standards for the most powerful AI systems and protects whistleblowers who disclose safety issues. Other laws target AI chatbots that induce minors to self-harm, require social platforms to add mental health warnings, and clarify that companies cannot waive liability for their algorithms or amplification of harassment. The author notes that California has not yet completed legislation on precise geolocation data protection, algorithmic fairness assessments, and financial accountability for platforms that harm children, and the 2026 session needs to continue advancing these.
