Back to home
AI Security
2 articles tagged with this topic
MCP ProtocolAnthropic
AI 工具互联的「插座标 准」MCP，正在从开发者 玩具变成企业级基础设施——但安全漏洞还没 补齐
MCP is graduating from developer toy to enterprise AI backbone, but critical security vulnerabilities haven 't been patched yet.
Apr 191 min read
AI SecurityPrompt Injection
Cross-Modal Prompt Injection Goes Open Source: The Security Mine Has Been Planted for Enterprise AI
A set of 23,759 cross-modal attack payloads has been made public, injecting fragmented text, images, PDFs, and audio into AI systems. Current single-c
Apr 103 min read