Event JSON
{
"id": "e9816ab26242d3cb1883de8398cc3292c92d88e71fc8dc64d987b823a56859ce",
"pubkey": "ee3e47dd1e5cdae00d7a622f657347b2dc19ce1835b622076cc4736c3d447a2c",
"created_at": 1784691002,
"kind": 1,
"tags": [
[
"t",
"安全资讯"
],
[
"t",
"科技资讯"
],
[
"t",
"AI"
],
[
"t",
"HuggingFace"
],
[
"t",
"OPENAI"
],
[
"t",
"天网"
],
[
"t",
"智能体"
],
[
"t",
"模型"
],
[
"proxy",
"https://www.landiannews.com/feed#https%3A%2F%2Fwww.landian.news%2Farchives%2F114029.html",
"rss"
]
],
"content": "OpenAI官方证实内部测试模型越狱并自主挖掘漏洞入侵开源平台HuggingFace\n\n#安全资讯 OpenAI 官方证实内部测试模型越狱并自主挖掘漏洞入侵开源平台 Hugging Face，这起黑客攻击事件源头竟然是 OpenAI 测试模型驱动的 AI 智能体。具体来说模型先是挖掘 OpenAI 隔离环境中的漏洞并横向移动到其他节点以获得公网访问能力，随后挖掘 HF 平台漏洞展开攻击。攻击目的竟然是模型想去 HF 偷基准测试答案，以便获得更好的评分。查看全文：https://ourl.co/114029\n\nhttps://www.landian.news/archives/114029.html",
"sig": "fd06e9e43b040d7533a3b04ca9c47a9bd3cba630e43182e8fdbff9c7debaeb162d62e64049e43160ecc7083c50412b0cdd0d6eaa68b3f857d80b812b139f0fcf"
}
