The recent rollout of OpenClaw has been a lightning rod moment for AI.
From a geek perspective? It's awesome. Watching an AI agent that can pretty much do anything on a computer with no guardrails - that's exciting to experiment with.
From a business perspective? That's running barefoot across a floor covered in razor blades. You are guaranteed to get cut.
I'm a believer that sometimes you have to run with scissors. But AI with absolutely no guardrails is not something we can deploy in business today.
What I'm watching for is when someone takes that capability and wraps intelligent walls around it. Limiting what systems it can access, what sites it can reach or what actions it can take.
The security challenge going forward isn't just traditional infosec - logins and access controls. It's capability management. What are we going to allow this thing to do?
If we give an AI access to email the same way we'd give a new employee access, it could theoretically send a message to every customer. That's not okay.
The exploratory phase is building great capabilities. But guardrails are essential before this is ready for production.