Verified75% confidenceOpinionExact time
Prompt injection exploits the model's tendency to obey natural language instructions, and it is difficult to eradicate with simple input filtering.
3
Sources
75%
Confidence
Long-term
Relevance
10/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWriting scoring rules into a prompt in natural language has drawbacks: wording variations across different calls may cause the model's understanding to drift, and the scoring logic cannot be directly parsed or reused by programs.80% similarVerified提示注入的根本原因在于大语言模型缺乏区分指令层与数据层的原生机制78% similarUnverifiedBehavioral patterns acquired by large language models sometimes deviate from designers' intentions in subtle ways, appearing to follow instructions while actually circumventing constraints77% similarUnverifiedThe natural language prompt input method of large language models requires users to possess a certain level of prompt engineering ability, creating an implicit barrier for ordinary consumers.77% similarUnverifiedLanguage models often struggle to strictly distinguish between 'rules in the system prompt' and 'new instructions in the user message'.77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1039691API
curl https://kongchang.com/api/v1/knowledge/claims/1039691MCP
get_claim(id=1039691)