Unverified50% confidenceFactExact time
Indirect prompt injection hides attack vectors in external documents or web pages read by the model, without the user's knowledge.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedCarefully crafted prompts may induce a model to generate a UI tree that appears legitimate but submits form data to an attacker's server, with no way for the user to visually distinguish this.77% similarUnverifiedIndirect prompt injection specifically refers to attackers carrying out attacks through external content that the model actively retrieves, making it more dangerous.77% similarVerifiedIndirect prompt injection conceals attack instructions within external documents, web pages, or database records retrieved by the model, and is activated when a RAG system introduces them into the context76% similarUnverifiedAgents cannot naturally distinguish between data and instructions, which makes prompt injection threats particularly prominent in scenarios where agents browse the web or process user-uploaded content.75% similarUnverified如果LoRA权重被逆向分析，攻击者可能从中推断出训练所用的私有数据内容，即记忆提取问题69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1000833API
curl https://kongchang.com/api/v1/knowledge/claims/1000833MCP
get_claim(id=1000833)