The AI Security Gap. The documents were access-controlled, but RAG embeddings aren't. The folder had permissions; the vector has coordinates. The fix isn't a better model, and it isn't a sternly worded system prompt. It's architecture: permissions that survive embedding, identity-aware filtering at retrieval time, and a security review that treats the vector store as what it is—a copy of your most sensitive knowledge with the traditional locks filed off.

AI agents are not employees; they are a different kind of participant with special requirements and risks.

AI agents and LLMs are probabilistic system and can hallucinate by design. It can't be stopped. Right architecture can manage it.

Threat of IA is overestimated. AI is not developed to be the primary risk yet. There are other more important and urgent risks and threats at the base of the society.