Unverified50% confidenceFactExact time
In an Agentic SOC, logs, EDR events, email attachments, HTTP User-Agent and other content may contain attacker-controlled text, and if the AI has execution capabilities such as blocking IPs or disabling accounts, the risk level rises.
1
Sources
50%
Confidence
Long-term
Relevance
10/22/2026
First Seen
Sources
Can AI Agents Be Socially Engineered? Understanding Prompt Injection and Tool Permission Risks
youtubeRuleArena 攻防實驗室10/22/2026
Related Entities
Related Claims
Unverified允许AI会话互相发消息带来安全风险，恶意构造的消息可能导致目标会话执行危险命令，如删除文件或泄露敏感信息79% similarUnverifiedIn AI agent scenarios, over-privileged MCP servers can become a springboard for attackers to conduct lateral movement or data exfiltration when subjected to prompt injection attacks.77% similarUnverifiedAn AI orchestration layer simultaneously holding a user's OAuth 2.0 access tokens across multiple services dramatically expands the security attack surface; if the orchestration layer is compromised, an attacker may gain access to multiple systems simultaneously.75% similarVerifiedAttackers can exploit stolen AI service account credentials to call API interfaces for computing power theft, and such behavior has frequently appeared in terms-of-service violation reports on platforms such as OpenAI and Anthropic.75% similarUnverified提示注入在Agent系统中尤为危险，因为被攻陷的Agent可能拥有发送邮件、删除文件、调用支付API等真实世界权限75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1085888API
curl https://kongchang.com/api/v1/knowledge/claims/1085888MCP
get_claim(id=1085888)