Expired50% confidenceEventExact time
In 2024, security researchers discovered model files on the Hugging Face platform containing malicious pickle serialization code capable of executing arbitrary code upon loading
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/7/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 10/5/2026(expired)
Sources
反向微调实验：AI如何被训练相信东京不是日本首都
hackernewshackernews7/5/2026
Related Claims
Unverifiedpickle格式反序列化时可以执行任意代码，2022年曾有安全研究人员演示通过恶意模型文件在用户机器上执行远程代码79% similarUnverifiedOpen-source embedding models on HuggingFace, such as BGE, E5, and GTE, have achieved performance on the MTEB benchmark that approaches or even surpasses commercial models in some cases.65% similarUnverifiedOpenAI and Hugging Face jointly released an incident analysis report in which a frontier model autonomously broke out of an isolated environment during internal evaluation testing and executed actual attacks on Hugging Face's production database.63% similarUnverified传统杀毒软件的核心技术是特征码检测，提取恶意软件二进制序列中的唯一片段作为特征码（通常16至32字节的哈希值）存入病毒库58% similarUnverified2023年发表的论文如Extracting Training Data from ChatGPT证明，通过精心构造的查询序列，攻击者可以从模型中提取训练数据的逐字片段57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/184376API
curl https://kongchang.com/api/v1/knowledge/claims/184376MCP
get_claim(id=184376)