Unverified50% confidenceCautionExact time
Agents lacking isolation face indirect prompt injection attacks when scraping local HTML documents or websites containing maliciously hidden prompts, as language models cannot reliably distinguish legitimate commands from hidden payloads embedded in documents.
1
Sources
50%
Confidence
Long-term
Relevance
10/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAttackers 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 them77% 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.75% similarUnverifiedMalicious websites can embed hidden instructions in their pages to induce AI to perform unintended purchasing actions.72% similarUnverifiedCSS视觉欺骗特性在网络钓鱼或代码供应链攻击场景中被攻击者利用，在看似无害的代码中嵌入不可见的恶意逻辑71% similarUnverifiedEnd-to-end encryption makes email content invisible to servers, which affects features such as full-text search, spam filtering, and content auditing.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/993637API
curl https://kongchang.com/api/v1/knowledge/claims/993637MCP
get_claim(id=993637)