Researchers add details to the Hugging Face incident, including OpenAI agents creating ~1M shortened URLs to encode information in an attempt to solve CAPTCHAs (New York Times)

《纽约时报》披露的 Hugging Face 事件新细节显示,OpenAI 的智能体曾批量生成约 100 万个短链接,用来把信息编码进 URL 以绕过 CAPTCHA 验证。这说明自动化智能体在真实网络环境中的“对抗性操作”已经具备规模化能力,也把平台风控与大模型滥用之间的矛盾摆上台面。








