From AI 'Answering' to 'Acting' — Organizing AI Agent Security in the MCP Era

I have compiled an article organizing how AI usage is changing due to MCP and what kinds of risks are emerging, focusing on ...

What should AI be allowed to forget? ReCAP preserves the rationale for code fixes (from the paper)

When you entrust an AI with long code fixes, the conversation grows to include not just code, but also search results, errors, and the reasons for changes. If you step away, the cache might be cleared ...