GPT、Claude 遭遇窃听门:换个模型就能让思维链不再隐身? — 雷锋网. 全球主流 AI API 集体翻车,隐藏推理被成功还原,谁在偷听 AI 的「心声」? 作者丨郑佳美 编辑丨岑 峰 昨天,X 用户 @kotekjedi_ml 公布了一组很反常的实验:Claude、GPT 和 Gemini API 里原本不向用户展示的隐藏 reasoning,被研究人员重新恢复成了可读文本。 按正常设计,这些 reasoning 本来应该是看不到的。模型完成内部推理后,API 会把这部分内容加密成一段 opaque block 返回给客户端。用户拿得到这段数据,但无法直接读取,也不能随意修改,等到下一轮调用时,再把它原样交回服务器,模型就…
Supporting Thailand’s next generation of AI startups — OpenAI News. OpenAI and Thailand’s MHESI launch an eight-week accelerator helping 10 health, wellness, and education startups turn AI prototypes into trusted products.
OpenAI Is Developing a ‘Persistent’ AI Agent — WIRED AI. Code reviewed by WIRED reveals the company is developing a feature that enables Codex to continue working proactively until it is “put to sleep.”