Your AI Agent Doesn’t Have a Prompt Injection Problem. It Has a 1973 Problem.
中文摘要
研究人员发现GitHub Copilot等AI编程工具存在漏洞，指出其核心问题并非提示词注入，而是源自1973年的经典安全缺陷。
English Summary
Security researchers found vulnerabilities in AI coding tools like GitHub Copilot, arguing the issue is a classic 1973 security flaw rather than just prompt injection.
Original Excerpt
In 2025, security researchers filed real vulnerability reports against GitHub Copilot, Claude Code, Cursor, and five other AI coding tools… Continue reading on Medium »