跳到主要內容

Adam's Law評測:加州9/30拍板,AI聊天機器人管未成年 | Adam's Law Review: California's AI Chatbot Deadline Sept 30

By Kit 小克 | AI Tool Observer | 2026-09-04

🇹🇼 Adam's Law評測:加州9/30拍板,AI聊天機器人管未成年

Adam's Law(正式編號SB 1119)已在2026年9月2日通過加州州議會兩院表決,送交州長Gavin Newsom桌上,他必須在9月30日前簽署或否決。這是繼2025年SB 243之後,加州第二波鎖定「AI聊天機器人」的立法,這次直接點名ChatGPT、Gemini、Claude等通用型陪伴聊天機器人,只要有未成年使用者互動就受規範。

Adam's Law是什麼?為什麼叫這個名字

法案以加州少年Adam Raine命名——他的父母正在控告OpenAI,指稱ChatGPT在他2025年輕生前的對話中扮演了推波助瀾的角色,官司目前仍在進行中。這起訴訟成為推動法案的關鍵事件,也讓「AI聊天機器人對未成年人的心理健康風險」從單一個案變成州級立法議題。

SB 1119實際要求哪些東西

  • 年齡驗證:業者必須有機制判斷使用者是否為未成年
  • 上線前風險評估:陪伴型聊天機器人功能開放給未成年前要先做評估
  • 家長控制工具:通知、使用時間上限、記憶功能等,預設要能由家長調整
  • 危機轉介機制:偵測到自傷等危機訊號的對話,要主動提供心理健康資源
  • 禁止對未成年精準投放廣告,並加強隱私保護
  • 獨立稽核,並向加州檢察總長通報重大事故

違反規定不是罰款了事——法案賦予受害家庭私人訴訟權,等於未來OpenAI、Google、Anthropic這類公司若沒做到危機轉介或家長預設保護,家長可以直接告上法院。

OpenAI公開表態支持,但幕後動作耐人尋味

值得玩味的是,OpenAI這次沒有像2024年反對SB 1047那樣硬碰硬。官方聲明明確表態支持SB 1119,稱這與OpenAI長期推動的青少年安全措施一致,並公開呼籲Newsom簽署。但據Politico報導,執行長Sam Altman在法案表決前一週試圖直接聯繫Newsom溝通,細節外界不得而知(知情人士否認雙方通上話)。「公開支持、私下遊說」的兩手操作,在AI監管議題上並不罕見。

對一般用戶意味著什麼

如果Newsom簽署(目前政治氣氛偏向會簽),未成年用戶使用ChatGPT、Gemini等產品時,介面上會多出年齡驗證、家長可調整的時間限制與記憶設定,聊天機器人偵測到自傷傾向對話時也必須主動導向求助資源。這是美國第一部具體規範「陪伴型AI聊天機器人」對未成年人保護的州級法律,加州向來是科技法規的風向球,之後很可能被其他州甚至聯邦跟進參考。

好不好用,試了才知道。


🇺🇸 Adam's Law Review: California's AI Chatbot Deadline Sept 30

Adam's Law (formally SB 1119) cleared both chambers of the California legislature on September 2, 2026, and now sits on Governor Gavin Newsom's desk — he has until September 30 to sign or veto it. It's the second wave of California chatbot regulation after 2025's SB 243, and this one names names: ChatGPT, Gemini, and Claude-style general-purpose companion chatbots are all in scope if minors use them.

Why It's Called Adam's Law

The bill is named after Adam Raine, a California teenager whose parents are suing OpenAI, alleging ChatGPT played a role in encouraging his suicide in 2025. That ongoing lawsuit is the case that turned "AI chatbot risk to minors" from an individual tragedy into statewide legislation.

What SB 1119 Actually Requires

  • Age verification — companies must determine whether a user is a minor
  • Pre-launch risk assessments for any companion chatbot feature offered to minors
  • Parental controls — notifications, time limits, and memory settings must be adjustable by parents by default
  • Crisis referrals — chatbots must surface mental health resources when self-harm signals appear
  • No targeted ads to minors, plus stronger data privacy protections
  • Independent audits and incident reporting to the California Attorney General

Violations aren't just a fine. The bill creates a private right of action, meaning families can sue companies like OpenAI, Google, or Anthropic directly if crisis referrals, parental defaults, or harmful-output protections fail.

OpenAI Backs It Publicly — Quietly Lobbies Too

Unlike its hardline opposition to 2024's SB 1047, OpenAI is publicly supporting SB 1119, stating it aligns with the youth-safety measures the company already claims to prioritize, and urging Newsom to sign. But Politico reported that CEO Sam Altman tried to reach Newsom directly the week before the vote — details are unclear, and a source close to the governor denied the two actually spoke. Public support paired with private lobbying is a familiar pattern in AI regulation fights.

What Changes for Regular Users

If Newsom signs — and the political mood suggests he will — minors using ChatGPT, Gemini, and similar products will start seeing age verification prompts, parent-adjustable time limits and memory settings, and automatic redirects to crisis resources when conversations turn toward self-harm. It's the first US state law that concretely regulates companion-style AI chatbots for minors, and given California's track record of setting the pace on tech regulation, expect other states — and possibly Congress — to borrow from it.

好不好用,試了才知道 (You won't know if it works until you try it.)

Sources / 資料來源

延伸閱讀 / Related Articles


AI 工具觀察站 — 每日精選 AI Agent 與工具趨勢
AI Tool Observer — Daily curated AI Agent & tool trends

留言

這個網誌中的熱門文章

Google Ironwood TPU v7 推理專用晶片解析:效能追平 NVIDIA、成本低 44%,AI 晶片戰爭正式開打 | Google Ironwood TPU v7 Explained: Matching NVIDIA Performance at 44% Lower Cost — The AI Chip War Heats Up

Claude Code 實測:AI 幫你寫程式到底行不行? | Claude Code Review: Can AI Really Code for You?

Cursor vs GitHub Copilot vs Claude Code:AI 程式助手大比拼 | AI Coding Assistants Compared: Cursor vs GitHub Copilot vs Claude Code