Unverified50% confidenceFactExact time
Indirect prompt injection does not require an attacker to have direct access to the agent; it only requires poisoning a data source that the agent will actively read.
1
Sources
50%
Confidence
Long-term
Relevance
10/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedStealthy false data injection refers to attackers carefully crafting tampered data to bypass conventional anomaly detection mechanisms.76% similarUnverifiedIndirect injection refers to attackers hiding malicious instructions in external content that the model will read, such as webpage body text, email body text, or document content.75% similarVerifiedIndirect prompt injection attacks plant hidden instructions within normal business data to induce Agents into executing unauthorized operations, representing a security risk at the semantic level rather than the code level.74% similarUnverified将敏感信息置于模型可访问范围之外是防御 prompt 注入类攻击的有效隔离措施73% similarUnverifiedAttackers can hide malicious instructions in web pages, documents, or even images, and when an agent reads this content it may mistake the instructions for legitimate commands and execute them73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/971188API
curl https://kongchang.com/api/v1/knowledge/claims/971188MCP
get_claim(id=971188)