An interview with Tim New on GuardBox AI.

CyberSecure: Tim, over the past few months, you’ve been introducing a new product, GuardBox AI, to benefit offices nationwide. Our first question is simple: What is it?

Tim New: GuardBox AI is a secure, self-contained AI server that works just like the AI tools you’re used to — ChatGPT, Claude, Microsoft Copilot, etc. — but none of the information fed into it leaves your benefit office. It’s a standalone AI server designed specifically for labor unions that lives on-site. It doesn’t use the cloud, and it’s not a shared platform.

CyberSecure: What problem are you trying to address with GuardBox AI?

Tim New: AI tools are everywhere, and staff in benefit offices across the nation are naturally using them to write emails faster, summarize documents, or make sense of complex information. The question isn’t, “Are employees using AI?” It’s “Are they using it in a controlled, prudent, and defensible way.” 

AI tools make it incredibly easy to leak sensitive information inadvertently: PII, ePHI, pension records, claims data, legal communications. Even well-intentioned staff can paste something into an AI chatbot without realizing that the information will leave the organization for their prompt to be processed. 

And it’s not just copy-and-paste. Someone may upload an Excel file, a PDF, or a large document to an AI chatbot for help without realizing it contains confidential information. AI makes it effortless to drop in a 100-page report — or even dozens of documents at once — without reviewing every page
in advance.

CyberSecure: Shouldn’t a robust AI policy keep these things from happening?

Tim: Yes, but writing a comprehensive AI policy is a lot more difficult than it might seem. Policy is a starting point, but fiduciary protection requires real safeguards. 

In the building trades, you wouldn’t send someone onto an active jobsite with nothing but a safety manual and a quick training and say, “Be careful.” Of course, you need rules. Of course, you need education. But you also need protective equipment — hard hats, harnesses, barriers — because the environment has to be built for safety, not just for the worker's intentions. AI is similar.

If the AI engine lives out on the open internet, then the office is always one innocent copy-and-paste away from sensitive information leaving the organization. When that happens, it’s not just a “technology mistake,” it’s a breach of fiduciary duty to protect member data prudently.

This is why benefit offices need something different than consumer AI. They need AI that fits the fiduciary environment. They need real guardrails in place, not just good intentions.

CyberSecure: And that’s where GuardBox AI comes in?

Tim: GuardBox AI is based on a simple idea: If benefit offices are going to use AI — and they will — then they need a version that keeps sensitive information within the office.

GuardBox is an office-contained AI engine. It’s designed to support staff productivity while keeping member data inside the organization’s controlled environment. AI on the internet is public. AI on GuardBox is private, and nothing you feed it will ever leave the building the GuardBox is physically located in. 

CyberSecure: What do you hope trustees take away from this?

Tim: That AI is not just a technology conversation, it’s a fiduciary responsibility conversation.

Trustees don’t need to ban AI. They need to govern it. Proper governance requires a clear policy and the right physical architecture to make that policy realistic.

CyberSecure: Let me ask some practical questions: Isn’t something like this expensive to do? Does everyone in the office need their own box?

Tim: Those are fair questions, and the answer to both is “no.” A basic GuardBox AI system supports about 10 to 12 people using it at the same time. From the staff perspective, using GuardBox AI feels familiar. They access it through a normal web browser, just like they would with ChatGPT or Gemini. The difference is where the information goes. Instead of prompts and documents going out to a public server somewhere on the internet, they stay private, inside the office, routed to the GuardBox system on-site.

The cost is also intentionally straightforward. The basic GuardBox AI runs about $380 per month total, including everything — that’s an introductory price as we bring it to benefit offices for the first time.

So, this isn’t about buying a gadget for each person. It’s about giving the office a governed AI environment that aligns with fiduciary responsibility.

An AI server that keeps your data in the box.

On-Site Intelligence. On-Site Security.

• No open internet exposure
• Double-firewalled, similar to layered network security
• Designed to prevent
• data leakage

Customized for Your Union.

• Trained on your union’s policies, practices, and documents
• Aligned with your office operations
• Optionally branded for your union

Click here for more information about Guardbox AI!