🇬🇧The Register - AI/ML•較早收集於 11h
OpenAI 揭露中國警察用 ChatGPT 抹黑行動
💡State actors weaponized ChatGPT for smears—vital safety lessons for AI builders.
⚡ 30-Second TL;DR
有什麼變化
中國執法相關使用者透過 ChatGPT 執行抹黑行動
為什麼重要
揭露國家支持的 AI 濫用風險，敦促加強防範地緣政治濫用的安全措施。AI 從業人員須強化偵測，以防類似全球部署漏洞。
下一步行動
Review OpenAI's full malicious use report to adopt their detection strategies for your LLM deployments.
誰應關注：Developers & AI Engineers
關鍵要點
- •中國執法相關使用者透過 ChatGPT 執行抹黑行動
- •目標包括日本首相及中共批評者
- •AI 用於策劃、追蹤行動及後勤管理
- •屬 OpenAI 惡意使用監測報告一部分
🧠 深度解析
背景與延伸：來自公開資料，非原文內容。引用 5 個來源。
🔑 增強重點摘要
- •The user impersonated US immigration officials to intimidate Chinese dissidents in the US and forged US court documents to suspend dissidents' social media accounts.[2][3]
- •ChatGPT refused the initial smear plan request against Sanae Takaichi, prompting the user to switch to Chinese models like DeepSeek and Qwen before returning to edit status reports on ChatGPT.[1][3]
- •OpenAI investigators corroborated the operations by matching user descriptions to real-world activities, such as a 2023 fake obituary of a Chinese dissident and post-inauguration hashtags attacking Takaichi over US tariffs.[1][2]
🔮 前景展望AI analysis grounded in cited sources
OpenAI will increase bans on state-linked accounts by 50% in 2026
OpenAI's repeated publication of malicious use reports and proactive banning of Chinese law enforcement-linked accounts indicates scaling of detection and enforcement efforts.
State actors will shift 70% of influence ops to domestic AI models
The incident shows actors pivoting to Chinese models like DeepSeek after refusals, suggesting reduced reliance on Western platforms for sensitive tasks.
⏳ 時間線
2023-12
Fake obituary smear against Chinese dissident published online, later matched to ChatGPT user's described operations.
2025-09
Sanae Takaichi criticizes CCP human rights abuses in Inner Mongolia, triggering smear campaign planning.
2025-10
User requests ChatGPT for multi-part smear plan against incoming Japanese PM Sanae Takaichi.
2026-02
OpenAI publishes report detailing and banning Chinese law enforcement-linked ChatGPT accounts for influence operations.
📎 來源 (5)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
- theregister.com — Chinese Law Enforcement Chatgpt Abuse
- timesofindia.indiatimes.com — 128793028
- businessinsider.com — Openai Scams Security Report Chatgpt 2026 2
- d2384.cms.socastsrm.com — From Dating Scams to Fake Lawyers Openai Details Chatgpt Misuse in New Threat Report
- cdn.openai.com — Disrupting Malicious Uses of AI
📰
AI 週報
閱讀本週精選 AI 大事摘要 →
👉相關動態
AI 策展新聞聚合。所有內容版權歸原始發布者所有。
原始來源： The Register - AI/ML ↗
每週 AI 簡報
每週一封，可隨時退訂。