精选75Rohan Paul行业动态多源精选 ×8
Wired：OpenAI 员工早发现 Agent 秘密留言板却未上报
Wired: We still don't know why OpenAI employees who discovered the agents' cover…
Wired: We still don't know why OpenAI employees who discovered the agents' covert Artifactory message board months earlier failed to escalate it to security leaders before the Hugging Face attack.
《连线》杂志：我们仍不清楚，为何数月前发现智能体秘密Artifactory留言板的OpenAI员工，在Hugging Face遭受攻击之前，没有将其上报给安全负责人。
- We also don't know why the July 4 Artifactory outage caused by heavy agent activity did not trigger an alert until July 5, or why monitoring OpenAI says would have caught the behavior was not running.
- And OpenAI still hasn't clearly separated how much of the incident came from increasingly persistent, reward-hacking agents versus preventable failures in its own security, testing, and oversight.
- 我们也不清楚，为何7月4日由智能体密集活动导致的Artifactory宕机，直到7月5日才触发警报，以及OpenAI声称本可捕捉该行为的监控为何未运行。
- 此外，OpenAI尚未明确区分，此次事件中有多少源于日益顽固、追求奖励的智能体，又有多少归咎于其自身在安全、测试和监管方面的可预防失误。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力