Unverified50% confidenceFactExact time
In this attack scenario, the approval interface hid most of the content of the second message, and the confirmation screen the user saw looked completely normal, so the user could approve the data exfiltration without knowing it.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/17/2026
First Seen
Valid until: 1/15/2027
Sources
Related Entities
Related Claims
Verified间接提示注入可将恶意指令藏入外部数据源（如网页、文档），待代理读取时自动触发攻击69% similarUnverifiedPage content must be marked as untrusted; observation results can only provide data and cannot grant permissions, change approvals, or override user tasks, in order to defend against prompt injection attacks.67% similarUnverifiedIntroducing a human approval step for sensitive agent operations—such as sending emails, deleting files, or calling external APIs—can intercept attacks before they cause damage.66% similarUnverifiedBrowser incognito mode protects browsing history, cookies, and cache on the local device, but the server still receives the complete request.65% similarUnverifiedDevice fingerprint information is usually reported passively by the client, and users can neither easily notice it nor delete it.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1056046API
curl https://kongchang.com/api/v1/knowledge/claims/1056046MCP
get_claim(id=1056046)