
会谈前,央视旗下账号「玉渊谭天」本月第二次就中美 AI 对话发声,这次逐条拆开 Anthropic 的隐私政策;这篇把它的每条指控拿去对 Anthropic 自己的公开文件,说清哪些对得上、哪些核不到。
"Anthropic PBC's privacy-policy revisions have increased risks to user data by allowing information to be shared with US intelligence agencies when the company considers it necessary without legal procedures, a social-media account affiliated with China Central Television said in a post on Saturday."
"We may use your Inputs and Outputs to train and improve Anthropic AI models, unless you opt out through your account settings. Even if you opt-out, we will use Inputs and Outputs for model improvement when: (i) your conversations are flagged for safety review to improve our ability to detect harmful content, enforce our policies, or advance AI safety research."
"We may share personal data with government authorities, law enforcement, or other third parties where, based on the information available to us, we have a good-faith belief that disclosure is reasonably necessary to (i) comply with applicable law, regulation or legal process, including for legal, tax or accounting purposes, or in response to an enforceable governmental request…"

"In each case, we disrupted the activity, used what we learned to strengthen our safeguards, and shared intelligence with authorities and industry partners, where appropriate."
"A clear distinction must be drawn between genuine security threats and mere technological competition."
| 文章的说法 | 现在能核到什么 |
|---|---|
| 三年修订隐私政策 13 次 | Anthropic 隐私中心列出 12 个版本的生效日期(2023-07-08 至 2026-07-08)与 7 篇修订说明;相差的一次取决于把 2025-10-08 那条只改生效日期的修订单算 7 |
| 数据默认用于训练,退出后仍可能被使用 | 现行隐私政策原文写明"除非你在账户设置里选择退出",退出的例外是"因安全审查被标记"的对话 6 |
| 数据来源"从 3 个增至 6 个" | 现行政策在训练数据下列出六类来源;"2024 年 6 月到 2025 年 9 月"这个时间窗本刊未核 6 |
| 可把用户数据交给美国情报机构 | 政策原文写的是"政府机构、执法机构",前提是公司善意相信披露确有必要,并列了四类用途;"情报机构""无需法律程序"是文章的表述 6 |
| 与情报机构共享技术指标 | Anthropic 9 月 10 日报告自述"与主管部门和行业伙伴共享情报";文章引的 2 月 23 日报告原文本刊未核 8 |
| 6 月 10 日致美国参议院信,2880 万次对话与 2.5 万个账号 | 未找到该信的公开文本 |
| 7 月三个"威胁情报经理"岗位及其语言与许可要求 | 招聘页已下线,无法核对 |
| 9 月 10 日报告分析约 2 亿次"蒸馏"交互 | 报告公开章节给出的是阿里巴巴关联方日均峰值近 300 万次交互、来自 3500 多个虚假账号 8 |
| 周日纽约会谈把 AI 安全列入议程 | 路透社引一名熟悉安排的人士;白宫与财政部没有公布议程文本 2 |
| Anthropic 对这篇文章的回应 | 截至本篇发稿,彭博社与洛杉矶时报的报道里都没有 Anthropic 就 9 月 19 日文章作出的回应 1 |
| 时间(北京) | 看什么 |
|---|---|
| 9 月 20 日 22:30 | 贝森特与何立峰在纽约摩根大通总部的会谈开场,路透社口径的议题含 AI 安全 2 |
| 9 月 24 日 | 特朗普在白宫接待习近平,AI 护栏是双方公开分歧之一 2 |
| 会谈之后 | 中美 AI 对话是否公布日期、中方由谁牵头 3 |
| 更长时间 | 商务部、外交部、网信办是否接住这个账号的说法——这是账号表态与官方立场的分界 3 |
finance.sina.com.cn
m.weibo.cn
anthropic.com
privacy.anthropic.com
news.bloomberglaw.com
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.