Unverified50% confidenceFactExact time
Large language models currently have no reliable built-in defenses capable of identifying and rejecting prompt injection in all situations.
1
Sources
50%
Confidence
Long-term
Relevance
10/5/2026
First Seen
Sources
Related Entities
Related Claims
Unverified确定性防护无法处理自然语言层面的隐晦风险，是其主要局限78% similarUnverifiedCurrent large language models lack reliable self-introspection capabilities.75% similarUnverifiedLarge language models that have undergone safety alignment training often exhibit over-refusal behavior, erroneously refusing instructions that are themselves harmless but superficially touch on safety-related topics.72% similarUnverifiedDeterministic guardrails can only handle fixed-pattern content that looks like sensitive information and cannot detect natural language risks where sentences are fluent but subtly guide users toward incorrect actions.71% similarVerifiedWithout tool-calling capability, large language models can only operate within the language space and cannot truly affect external systems.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/978489API
curl https://kongchang.com/api/v1/knowledge/claims/978489MCP
get_claim(id=978489)