Unverified50% confidenceSolutionExact time
应在不可信内容到达分析模型之前通过清洗层处理，剥离或中和违规片段同时保留脚本其余部分供分析，而不是依赖模型实时自行判断
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
GuardBreaker攻击：用核武器提示词瘫痪AI恶意软件分析
Related Entities
Cite This Claim
Stable URI
https://kongchang.com/claim/1000246API
curl https://kongchang.com/api/v1/knowledge/claims/1000246MCP
get_claim(id=1000246)