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
IMF Confirms El Salvador’s Bitcoin Growth Was Funded by Private Donations
The International Monetary Fund said documentation shows El Salvador's bitcoin added since its first...
blockchainreporter2026-09-06 17:00:00
Connecticut Attorney General Warns Against Unregulated Offshore DeFi Exchanges
Connecticut Attorney General William Tong issued a consumer alert on offshore decentralized finance ...
blockchainreporter2026-09-06 15:00:00
Bank of Korea: Binance Stablecoin Pairs Can Push Local Currencies Lower
A Bank of Korea issue note finds that stablecoin pairs on Binance let demand for dollar-pegged token...
blockchainreporter2026-09-06 13:00:00