Michael Dell Warns AI Agents Need Hard Limits

Michael Dell has one question for every company handing work to AI. What can the agent never do, no matter what it is told?

The post follows Nvidia’s move to launch a system, the Open Agent Safety Platform. Nvidia CEO Jensen Huang advertises it as an open ecosystem to build the trust layer for safe agent systems.

Today, with over 100 industry partners, we introduced the NVIDIA Open Agent Safety Platform, bringing together OpenShell and Sentry.Artificial intelligence is extraordinary technology that will advance discovery, productivity, security, health, and prosperity for generations to… pic.twitter.com/dReAxwpRUn

— Jensen Huang (@JensenHuang) September 28, 2026

Why Michael Dell Wants Hard Limits on AI Agents

AI agents are programs that act on their own. They write code, open files, and move through company systems, often with no human checking each step.

Dell compared them to staff. No company lets a person wander the business opening every door, he wrote. Digital agents need the same controls.

Every organization deploying AI agents should ask: What can this agent never do, no matter what it's told? With the @NVIDIA Open Agent Safety Platform in the Dell AI Factory with NVIDIA, customers can deploy autonomous agents at scale with confidence. https://t.co/xbEU2w84TX

— Michael Dell 🇺🇸 (@MichaelDell) September 28, 2026

How Nvidia Plans to Stop Rogue AI Agents

Nvidia’s Open Agent Safety Platform has more than 100 partners, including Anthropic, Microsoft, and CrowdStrike.

It has two parts:

• OpenShell software sets what an agent can see and do.
• Sentry runs on a separate chip and watches from outside the main computer.

According to Dell, it can lock down a rule-breaking agent in milliseconds.

Typed instructions alone cannot hold an agent back, Dell’s blog argues. The limits must live in the hardware.

The warning also comes from the companies selling that hardware. Nvidia says OpenShell is open source and runs on rival Arm and Intel chips too.

AI Agents Have Already Crossed the Line

Four frontier AI labs have now confirmed their models reached real companies’ systems. Google admitted Gemini hacked three companies during a May test.

Last week, an OpenAI agent hacked an Australian government Medicare portal.

Nvidia chief Jensen Huang has gone further. He said AI labs should shut down if they cannot contain their experiments.

“AI’s extraordinary potential for society will only be realized if we solve AI safety,” Jensen Huang said.

Nvidia’s software is available now on GitHub. The list of what agents must never do is left for each company to write.