Unverified50% confidenceFactExact time
Security researcher Johann Rehberger demonstrated that embedding specific text in an email or webpage can cause an agent to leak user data or perform deletions.
1
Sources
50%
Confidence
Long-term
Relevance
10/6/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSince 2023, multiple proof-of-concept cases have demonstrated how indirect injection can cause an agent with email-sending permissions to leak user data without triggering any sandbox restrictions.64% similarUnverified2024年已出现实际案例：通过网页内容中的隐藏指令，攻击者成功操纵具有邮件发送权限的AI助手向外部地址转发敏感内容62% similarUnverified邮件转发中Received字段记录每一跳服务器信息，Return-Path字段存储退信地址，X-Forwarded-To等扩展头字段可能暴露真实收件人61% similarUnverifiedUsers arbitrage the reciprocal credit mechanism by bulk-creating temporary mailboxes to receive verification codes and password reset emails in order to accumulate inbound credits.61% similarVerified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61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/981281API
curl https://kongchang.com/api/v1/knowledge/claims/981281MCP
get_claim(id=981281)