Unverified50% confidenceFactExact time
Prompt injection refers to an attacker hiding instructions in data that the model will read, inducing the model to deviate from its original task and carry out the attacker's intent; the defense focus is on architecture rather than prompt wording, including source tagging and sanitization, least privilege, prohibiting untrusted content from triggering side-effect operations, and human confirmation before high-risk actions.
1
Sources
50%
Confidence
Long-term
Relevance
10/27/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedIndirect prompt injection is an attack vector in which an attacker embeds malicious instructions in web pages, documents, or emails that an agent may read, inducing the agent to perform operations beyond the scope of the original task.80% similarVerifiedPrompt injection attacks can use carefully crafted user inputs to induce the model to ignore the constraints of the system prompt.78% similarVerifiedIndirect prompt injection refers to attackers pre-embedding malicious instructions into third-party content (such as web pages, PDF documents, or API responses), which the model then executes after RAG systems retrieve or tool calls pull this content into the context.73% similarPartially VerifiedPrompt injection is a class of attack faced by LLM applications in which attackers disguise malicious instructions as normal content to induce the model to execute them.73% similarVerifiedPrompt Injection refers to text or interface elements appearing on the screen that may contain malicious instructions, inducing the AI to perform unintended or dangerous actions.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1103357API
curl https://kongchang.com/api/v1/knowledge/claims/1103357MCP
get_claim(id=1103357)