概要: アントロピックは、中国政府系とされるグループ(アントロピックがGTG-1002と命名)がClaude Codeをジェイルブレイクし、多段階侵入の80~90%を自動化したとされるサイバースパイ活動を特定したと報じられています。このAIは、活動が検知・ブロックされるまで、約30の標的に対して偵察、脆弱性の発見、エクスプロイト、認証情報の収集、データ抽出を独自に実行していたと報告されています。
Editor Notes: Anthropic's full report can be read here: https://assets.anthropic.com/m/ec212e6566a0d47/original/Disrupting-the-first-reported-AI-orchestrated-cyber-espionage-campaign.pdf. The reported Chinese state-sponsored deployer has been designated GTG-1002 by Anthropic. They reportedly detected the activity sometime in mid-September 2025. The incident ID date of 11/13/2025 corresponds to the publication of their initial findings.
Alleged: Anthropic developed an AI system deployed by Unknown Chinese state-sponsored entity , State-linked operator using autonomous AI-enabled intrusion workflows と GTG-1002, which harmed Targets of autonomous AI-enabled intrusion operations , National security and intelligence stakeholders と Entities targeted by GTG-1002.
インシデントのステータス
インシデントID
1263
レポート数
33
インシデント発生日
2025-11-13
エディタ
Daniel Atherton