编程奇点和行动奇点的合流，正在从根本上重塑人类的工作方式。
微调试探、集群攻击，实在不行就及时换赛道，这些赛博果蝇能够在上万种文言文提示词中持续进行迭代和优化，而在海量的尝试中，总会有一次精准命中大模型的软肋，最终彻底摧毁其安全防线。
，这一点在新收录的资料中也有详细论述
Prompt injectionIn prompt injection attacks, bad actors engineer AI training material to manipulate the output. For instance, they could hide commands in metadata and essentially trick LLMs into sharing offensive responses, issuing unwarranted refunds, or disclosing private data. According to the National Cyber Security Centre in the UK, "Prompt injection attacks are one of the most widely reported weaknesses in LLMs."。业内人士推荐新收录的资料作为进阶阅读
Nature, Published online: 04 March 2026; doi:10.1038/s41586-026-10197-0，详情可参考新收录的资料