0.01欧元转账可危及银行AI代理系统，bunq修复安全漏洞

2026-06-11 02:49·22天前·tvissers

AI 摘要

安全公司blue41发现，通过向bunq银行AI代理发送一笔0.01欧元的转账，攻击者可能利用提示词注入等漏洞操控AI代理，进而获取用户数据或执行未授权操作。该漏洞源于AI代理对用户输入与系统指令的边界处理不足。bunq已与blue41合作修复问题，并公开致谢。此次事件凸显金融AI智能体面临的新型安全风险。

原文 · 未翻译

🚀 Blue41 wins RSAC Launch Pad. Read more here.

Risk control platform for agentic AI

Enabling enterprises to deploy secure AI agents.

AI security expertise shared with leading organizations

Understand and control how your agents behave

See what data agents access, which actions they take, and which controls are working.

Detect and respond to agentic AI incidents.

Blue41 detects runtime incidents such as anomalous behavior resulting from prompt injection attacks and unauthorized agent activity.

Your partner for AI risk control.

Blue41 helps enterprises navigate the rapidly evolving risk landscape around AI agents.

Runtime risk platform

Risk monitoring of AI agents in production. Detect security incidents and compliance violations in real time.

Deep AI risk expertise

Expert support to understand emerging threats, assess architectures, and define controls for secure and compliant deployments.

Close collaboration

Ongoing partnership with your security, AI, and governance teams. From assessments to incident response.

Start the conversation

Blue41 helps enterprises assess, monitor, and control AI agent risk.

Hacker News 热门（buzzing.cc 中文翻译）

19导出 Markdown