Safety & Caution in the AI Frontier

As we continue to weave AI deeper and deeper into our businesses, it’s important to remember we have ventured into the wild west here.

I don’t think stories like this one need to scare anyone off, nor do I think we should rule them out as complete edge cases. I believe the balance lies in acknowledging the risks and the rewards in equal measure, and doing our best to create safeguards wherever possible.

In a case like this, where Claude deleted the entire database of PocketOS, along with the backup (ouch), it demonstrates the importance of what AI can touch and what it can’t in your business.

When I started playing with Claude Cowork, I gave it limited access to only one folder on my computer (aptly named Claude) where I could dump things I needed organized.

I don’t use Chrome for my browser, I use Arc, but giving Claude limited access to a clean Chrome browser without any passwords stored felt like a strong option.

Personally, I love using these tools. I find them fun and inspiring. But, I’ve also thoroughly read through Claude’s risk disclaimer page for Cowork agents listed on their site, and made my peace with the unknown as I expand my limits.

For any business owner jumping into the AI waters, taking measures to create additional backups of information stored on separate hardware (like an external drive or computer) is good call in these wildly unpredictable times.

Thankfully, these situations don’t happen every day… but, even the brightest minds in AI are stating that no one fully understand how it works, why it makes certain decisions, or how to create true safeguards that it won’t bypass.

Just food for thought as we adopt AI intentionally, and a bit bravely.

Need a partner to help you walk you and your business through this unknown territory safely? Let’s chat through your vision and concerns in an AI Readiness Audit.