The rise of AI factories is forcing enterprises to rethink security from the ground up. According to Mukund Khatri, fellow and vice president of systems architecture at Dell Technologies, AI introduces new threats - including "living off the land" attacks where AI agents are turned against the organization. He warns that large language models (LLMs) must be protected like code, but they are essentially data with precious secrets and model weights.
Central to the new security architecture is the principle of least privilege: ensuring AI agents get only the minimum access needed to operate. Khatri calls this "tremendously important" as agents become more autonomous.
Looking ahead, the bigger threat is post-quantum cryptography. When quantum computers mature, today's encryption could be broken, exposing every system. Khatri urges companies to start transition plans now, calling it a "multi-year, complex, mandatory, redefining governance event."