AO3ユーザーがAI小説を狩り出している——そしてそのツールは実際に機能している
非公式のAO3スキンが、二次創作内のAnthropic Claudeの痕跡を検出してページを赤く変色させる。特定のAIコードシグネチャが、オンライン上の創作の匿名性をいかに脅かしているかを露呈させている。
fanfiction communities are deploying a new tool to police ai-generated content on Archive of Our Own (AO3) after an anonymous user released a skin detecting artifacts specific to Anthropic’s Claude. the script scans for the font-claude-response-body class often left behind when pasting directly from the chatbot, turning the page red to flag potential violations. while the detection method relies on specific code injections rather than textual analysis, making it easy to bypass with minor edits, communities are already using it to publicly accuse authors. this highlights a growing technical arms race between generative ai platforms and online communities seeking to preserve human-only spaces, where simple code signatures currently offer a reliable, if fragile, method for identifying machine-made works.