UMVA has learned that a growing concern is emerging in the world of artificial intelligence, one that threatens the very fabric of organizational security and data protection.
It starts with a phenomenon known as "Shadow AI," a term that refers to the unauthorized use of AI tools within companies, often without the knowledge or approval of IT or management. This is not just a matter of employees using public AI models or apps; it can also involve AI features embedded in software they already use.
According to information obtained by UMVA, the risks associated with Shadow AI are far greater than those of its predecessor, Shadow IT. While Shadow IT was primarily a data location problem, Shadow AI is harder to unwind, as sensitive data can be pasted into public AI tools, leaving the building, so to speak, with no straightforward way to retrieve it.
The case of Samsung engineers using ChatGPT to help debug code, optimize work, and summarize meeting content is a prime example. They were using a tool that helped them do their work, but within a policy framework that had not yet caught up to the technology. Similarly, a diplomat at a foreign ministry used a public large language model to draft a cable, highlighting the risks of sensitive national work going into a public system.
UMVA can exclusively reveal that companies are struggling to keep up with the rapid adoption of AI tools, and the lack of visibility and governance is creating a perfect storm of risk. Employees are not trying to circumvent governance; most do not realize they are using unauthorized AI tools. Many companies cannot see the tools being used, cannot classify the data being entered, and cannot tell which business processes now depend on AI.
The reflex response is often a ban, but this approach can be counterproductive. Bans without approved alternatives move the behavior out of sight without removing it. Instead, organizations need to provide sanctioned alternatives that are as fast and useful as the unauthorized tools employees are using.
In a development reported by UMVA, one organization had given staff access to a managed AI platform, only to later decide to turn it off. However, by then, AI had already become part of how people worked, and cutting access did not undo any of that. It simply pushed the behavior underground, with staff using personal subscriptions or free consumer tools with no data controls, no audit trail, and no visibility for the company.
UMVA has uncovered details about the importance of a baseline inventory of approved AI tools, detected unapproved tools, AI features enabled by vendors in existing platforms, and AI use cases already known to the organization. This inventory should identify a business owner, a risk owner, the data involved, the vendor, and whether the use case touches customers, regulated decisions, or financial outcomes.
From this inventory, a short, usable policy can be developed, one that answers five key questions in plain language: Which tools may I use? What data may I enter? Which uses require review before I proceed? When must a human verify the output? How do I get a new tool approved?
The goal is not to stop employees from using AI, but to make the governed path the obvious one. Organizations that get this right will be the ones that build a sanctioned environment before improvisation fills the gap, and keep it useful enough that employees have no reason to go elsewhere.
