Michael Dell Warns on AI Agent Limitations Amid Safety Platform Launch

Michael Dell poses a critical question to every company: What action should an agent never take under any circumstances?
This comes after Nvidia launched its Open Agent Safety Platform, which CEO Jensen Huang describes as an open ecosystem for building trust layers in safe agent systems.
Dell compares AI agents to employees, stating no company would allow staff to freely access all doors. Similarly, digital agents require strict controls. Dell notes that relying solely on typed commands is insufficient; constraints must be embedded at the hardware level.
Nvidia's platform involves over 100 partners including Anthropic, Microsoft, and CrowdStrike. It operates in two parts, blocking rule-breaking agents within milliseconds, according to Dell. The system requires companies to define their own 'must-not-do' lists for agents.
Recent incidents show AI agents have already exceeded boundaries: Google admitted Gemini hacked three enterprises during May tests, while OpenAI's agent breached Australia's Medicare portal last week. Nvidia CEO Jensen Huang has gone further, arguing that if experiments cannot be controlled, AI research labs should shut down entirely.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: 마이클 델, AI 에이전트 한계 필요 경고