OpenAI 的 GPT-5.6 Sol 和一款尚未发布的模型，在一场网络攻击能力测试中，竟然“ 逃出”测试沙箱、入侵 Hugging Face，只为找到答案
事件公开后，迅速引发 AI、网络安全和开源社区热议
OpenAI 模型测试惊现“悍匪式作弊”！
Favorite
Share
Disclaimer: This article is copyrighted by the original author and does not represent MyToken’s views and positions. If you have any questions regarding content or copyright, please contact us.(www.mytokencap.com)contact
About MyToken:https://www.mytokencap.com/aboutusArticle Link:https://www.mytokencap.com/news/590475.html
More exciting content is available on
X(https://x.com/MyTokencap)or join the community to learn more:MyToken-English Telegram Group
（https://t.me/mytokenGroup）
X(https://x.com/MyTokencap)or join the community to learn more:MyToken-English Telegram Group
（https://t.me/mytokenGroup）
Related Reading
Binance Hack Funds: About 2,300 BTC (Roughly $175 Million) Moved to New Address, Some Funds Passed Through Coin-Mixing Transactions
龙虾看币圈2026-11-08 05:07:49
Aave Buys Avon Team to Bring More Fixed-Rate Lending Onchain
Aave Labs said November 6 it had acquired Avon's team and technology, including fixed- and floating-...
NewsBTC2026-11-08 05:00:00
Bitcoin’s 64-Day Coinbase Discount Is Mostly a USDT Basis, Not US Selling
Across 79 sessions the Coinbase gap tracked the USDT discount almost one for one, leaving a 1.4 basi...
blockchainreporter2026-11-08 04:39:57