Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

Why ‘free’ AI will cost you everything

Your DIY OpenClaw deployment is a ticking time bomb and here is how you can avoid malicious attacks.

[Photo: Getty Images]

Profile photo of Chongwei Chen

President and CEO · DataNumen, Inc.

As AI popularity continues to soar, employees in many organizations have taken matters into their own hands. Instead of waiting for their IT teams to enable approved AI systems, they are using free or open-source AI tools without a shred of worry. Often, the usage goes unnoticed or is seen as experimentation.

While some leaders take a lenient view of unapproved use of free AI tools, such practices expose organizations to serious risks, including data leaks and malicious prompt injection attacks.

A better way to adapt to free AI tools is to use their domesticated versions. The analogy comes from ecology: wild species are difficult to handle, say, a wild cat, whereas a domesticated breed like the house cat displays better temperament and has less destructive tendencies.

Domesticated versions of OpenClaw, such as NanoClaw, include a defined structure and sandbox features. With a domesticated AI version, an organization can maintain proper oversight while providing employees with the desired AI features.

WHY FREE AI IS A DISASTER IN THE MAKING

Open-source AI tools like OpenClaw have an undeniable appeal; not only are they free, but the setup process is simple and non-technical. Teams struggling with a business problem, or executives struggling with a data analysis problem, can download the tool and ask it to go to work. Much of this can occur under the radar, and IT teams may be unaware of its widespread usae.

Even if a business leader recognizes the risk of unauthorized usage, many may give exceptions due to the overall AI craze and the promise of outstanding results. However, raw OpenClaw deployments typically lack enterprise-grade security measures. They also lack sandbox environments and prompt injection checks. When such tools are self-deployed, granular controls are hardly applied. This creates a large attack surface for malicious actors, especially if the tool is widely used within the organization.

Given that the free AI tool was deployed without a vendor in place, or without the IT team in the loop, the security liability lands squarely on the organization. You never know when data will be leaked or when an AI agent will take an action that compromises your operations. OpenClaw and many other AI agents operate at an elevated authorization level, which can create a significant mess if they operate without supervision and safeguards.

FREE AI SHOULD BE A C-SUITE PRIORITY

Even a cursory cost-benefit analysis of a wild AI tool deployment can showcase the stark difference between savings and risks. While you may save a few thousand dollars across the organization by opting for a free AI tool, the typical costs associated with a data breach can run into millions. With a strict compliance environment in place, regulatory actions following a data breach can severely damage an organization’s financial health and reputation.

The issue at hand is not just a procurement decision but risk management, and it cannot be left to the IT team. In fact, AI governance squarely falls to C-suite executives, and they cannot shy away from taking the bull by its horns. 

Leaders need to implement organization-wide mandates against unauthorized use of free tools. At the same time, leaders need to have an open discussion with employees and enable AI alternatives that meet their needs.

Enterprise-friendly productized versions of OpenClaw exist because they address the gap between deploying a tool and running it securely. Business leaders can swiftly onboard such tools and curb unauthorized use of wild AI tools.

KEY BENEFITS OF ENABLING GOVERNANCE

There are several advantages to enabling a governance structure for open-source AI deployments by using productized versions. To start with, such deployments come with sandbox environments where specific failure points can be contained. Another core advantage that enterprise customizations enable is end-to-end encryption. This prevents proprietary data from being accessed by unauthorized third-party platforms with which the AI may interact. Moreover, enterprise implementations are also designed to handle prompt injection attacks, detect anomalies, and raise alerts. 

Governance also enables the implementation of granular controls and the restriction of access to AI agents for sensitive data. It can also restrict certain actions, like deleting a CRM record or dropping a database without human go-ahead.

FINAL THOUGHTS

The lure of free AI tools and apparent ease of deployment can compromise the judgment of many experienced leaders. However, it’s prudent to understand the nature of the beast; AI is difficult to govern in the best of scenarios, and a wild AI tool floating within an organization is a direct path to a calamity. 

Leaders need to strike a balance and take advantage of AI tools like OpenClaw by using a domesticated version with proper governance controls and safety paradigms in place.

Chongwei Chen is president and CEO of DataNumen.