Unverified50% confidenceFactExact time
Direct prompt injection occurs in the conversation between the user and the model, and its impact is usually limited to that user's own session; with indirect injection the victim and the attacker are often not the same person, making it more covert.
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWhen an interaction itself contains no obviously violating content, it is difficult for a model to determine the user's true intent71% similarUnverifiedIndirect prompt injection is more harmful because the victim is completely unaware of the malicious content.67% similarUnverifiedWhen the same model both writes and reviews code, it tends to be overly lenient toward its own output; therefore the critic and the fixer need to be played by different Agents in independent sessions.67% similarUnverifiedSession affinity binds a session to a single provider, so when that provider experiences fluctuations, the continuity of the session may be affected.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1066729API
curl https://kongchang.com/api/v1/knowledge/claims/1066729MCP
get_claim(id=1066729)