[KongchangAI]
Unverified50% confidenceOpinionExact time

The original poster's developer acknowledges that the boundary marker scheme is only a mitigation measure, not a security guarantee, and that the model may still be bypassed past the boundary markers by a clever prompt injection.

1
Sources
50%
Confidence
Long-term
Relevance
11/3/2026
First Seen

Sources

Related Entities

Related Claims

Cite This Claim

Stable URI
https://kongchang.com/claim/1132375
API
curl https://kongchang.com/api/v1/knowledge/claims/1132375
MCP
get_claim(id=1132375)