Kief Studio

AI security

Notes

AI security

Every Kief Studio note filed under AI security, newest first. Read one before you decide to write to us.

  1. Protect

    Voice-Activated Browsing: The Real Risks of Talking to an AI Browser

    Voice-activated browsing risks backed by research and regulators: inaudible commands, cloned voices, prompt injection and stored recordings, plus a one-hour fix list.

    Agentic Browsers / Voice assistants / Voice cloning

    Brian Gagne

  2. Protect

    Prompt Injection Attacks Explained: How Massachusetts Users Can Stay Protected

    How prompt injection tricks AI browsers and assistants, why it can't simply be patched, and the steps Massachusetts users and businesses can take today.

    Brian Gagne

  3. Protect

    The Rise of Agentic Malware: Protecting Your Data Before It Thinks for You

    Agentic malware asks an AI model what to do mid attack. The documented cases from 2025, what is still hype, and six practical defenses for a small business.

    Brian Gagne

  4. Automate

    Beyond the Hype: How We Build Multi-Model AI Agents That Hold Up

    Multi-model AI agents swap models without a rewrite. Vendor retirement policies, OWASP's excessive agency risk, and the 7-step build order we follow.

    Brian Gagne