Artificial Intelligence (AI)
AI Agents Can Be Secured. We Can Do It.
Learn of the benefits of safely securing AI agents.
Save to Folio
I strongly support the industry’s efforts to build safety and security into AI agents.
After decades as an IT and cybersecurity engineer and CEO, I have had the privilege of witnessing many waves of technological change.
And I have seen the same pattern repeatedly:
New technology creates new vulnerabilities.
New vulnerabilities create fear.
And engineers eventually learn how to secure the technology and make it useful at scale.
AI and AI agents are going through that same cycle today.
In fact, the debate reminds me of the early days of electricity — the famous “current war.”
Electricity was invisible, enormously powerful, and potentially deadly. People had legitimate reasons to be afraid.
But imagine if humanity had said:
“Electricity is too dangerous. Stop.”
There would be no modern electrical infrastructure, no computers, no Internet, no modern hospitals, and no technological civilization as we know it.
Instead, engineers built power plants, transmission systems, insulation, grounding, circuit breakers, safety standards and countless devices designed to use electricity safely.
The solution was not to stop electricity.
The solution was to engineer safety around it.
I believe AI is similar.
AI is becoming something like electricity for intelligence — enormously powerful, invisible as software, and capable of flowing into almost every aspect of human life.
But the challenge is not only the AI model.
It is the infrastructure around the model.
How do we test it?
How do we measure its behavior?
How do we authenticate it?
How do we control its permissions?
How do we isolate it?
How do we monitor its actions?
How do we create governance and accountability?
And, most importantly:
How do we stop an agent when it crosses a boundary?
These are cybersecurity questions.
We may never be able to control how intelligence thinks — whether human intelligence or artificial intelligence.
As a philosophy major, I still love Descartes:
“I think, therefore I am.”
我思故我在。
Human civilization has progressed precisely because we have never completely controlled human thinking.
We think.
We question.
We imagine.
We innovate.
But civilization has also learned something equally important:
We don’t have to control how people think in order to govern how people act.
That is exactly where cybersecurity can contribute to the AI era.
We can build the guardrails.
We can build identity and authorization.
We can build isolation and monitoring.
We can build safety testing.
We can build policy controls.
We can build auditability.
And we can build stop gates into AI agents.
In other words:
We don’t necessarily need to control the intelligence.
We need to engineer how that intelligence is allowed to act.
That is what cybersecurity engineers have been doing for decades.
We have done it before.
And I believe we can do it again.
AI has enormous potential to help humanity deal with some of our greatest challenges — healthcare and disease, climate and weather, scientific discovery, education, productivity, and preparing the next generation for a rapidly changing world.
We should not stop because AI is powerful.
We should make it safe because it is powerful.
This is why I support the industry effort to make AI agents secure, governed and controllable.
Cybersecurity engineers have an important role to play in this next chapter.
We are not here to build walls around intelligence.
We are here to build the bridges that allow intelligence to safely enter the real world.
Like the old song says:
“Like a bridge over troubled water…”
Let us build the bridges.
Let us build the safeguards.
Let us build the governance.
Let us build the safety agents.
And let us help humanity use this extraordinary new source of intelligence responsibly.
AI agents can be secured.
We can do it.