Nobody sends an email announcing they have started pasting company data into ChatGPT. Shadow AI spreads quietly, driven by people who are just trying to get their work done faster. By the time someone in leadership notices, the usage is widespread and the data exposure is already real.
According to IBM's 2025 Cost of Data Breach Report, one in five organizations has already experienced a breach linked to unsanctioned AI. Shadow AI incidents add $670,000 to average breach costs. And research from Harmonic Security found 665 distinct generative AI tools operating across enterprise environments, with only 40% of companies having purchased an official AI subscription.
The question is not whether your team is using unauthorized AI. It is whether you know about it. Here are five signs that suggest you do not.
Sign 1: Output quality improved but nobody changed the process
A team that was producing mediocre email drafts, reports, or proposals suddenly starts producing polished ones. The turnaround time dropped. The writing sounds more confident. The formatting is suspiciously consistent.
Nobody attended a training. Nobody hired a new person. Nobody changed the template. Someone on the team found an AI tool that makes their work better, and they are using it without telling anyone.
This is the most common and the least alarming sign of shadow AI. The risk is not the quality improvement. The risk is what data the team member is feeding into the AI tool to get those results. If the input includes customer names, deal values, internal strategy documents, or employee performance data, you have a data exposure problem wrapped in a productivity win.
Sign 2: New AI-related charges on expense reports or credit cards
Check your expense reports for subscriptions to tools like ChatGPT Plus, Claude Pro, Jasper, Copy.ai, Perplexity, or any of the hundreds of AI products that charge $20-50 per month. Check corporate credit card statements for recurring charges from OpenAI, Anthropic, or similar providers.
If employees are paying for AI tools out of pocket or on corporate cards without going through procurement, that is shadow AI by definition. The tool has not been vetted for security, data handling, or compliance.
This sign is easy to detect and easy to address. The harder version is when employees use free-tier AI tools, which leave no financial footprint at all.
Sign 3: Browser extensions you did not approve
AI-powered browser extensions are one of the fastest-growing vectors for shadow AI. An employee installs an extension that summarizes emails, drafts replies, or extracts data from web pages. The extension requires broad permissions ("read and change all your data on all websites") and the employee clicks "Allow" without reading the fine print.
Now every page that employee views, including internal dashboards, CRM records, and financial reports, is potentially accessible to the extension's third-party backend. The employee sees a productivity tool. Your security team would see a data exfiltration risk.
As one Repello AI analysis noted, browser extensions, Slack bots, and CI/CD pipeline scripts are shadow AI hiding spots that traditional SaaS discovery tools are not designed to detect.
Sign 4: People are getting answers faster than your systems can provide them
An ops manager who used to spend two hours compiling a weekly report now delivers it in 20 minutes. A support agent who used to take 15 minutes to draft a complex response now sends it in three. A finance analyst who struggled with a reconciliation process now breezes through it.
Speed improvements this dramatic rarely come from working harder. They come from working with AI. If your sanctioned toolset has not changed, the AI assistance is coming from somewhere unsanctioned.
This is the sign that should trigger a conversation, not a crackdown. The employee has identified a real productivity gain. Your job is to provide a sanctioned way to capture that gain without the data risk.
Sign 5: The phrase "I asked AI" has become normal and nobody defined the rules
Listen to how your team talks about their work. "I ran this through AI." "AI helped me draft this." "I asked AI to summarize the meeting notes." If these phrases are part of casual conversation but your organization has no AI usage policy, guidelines, or approved tools, you have shadow AI by default.
The absence of a policy is itself a risk. When there are no rules, every employee makes their own judgment about what data is safe to share with AI tools. Some of those judgments will be wrong.
For a deeper dive into the full scope of this problem, read our comprehensive guide on what shadow AI is and how to address it.
What to do when you see these signs
The research is clear on one point: banning AI tools does not work. Microsoft's Work Trend Index found that 78% of AI users bring their own tools to work, and a separate study from DNV found that workers in over 90% of organizations that banned AI continued using personal AI tools for work tasks.
Bans push usage underground. Governance brings it into the light.
The fix is three steps:
-
Audit current usage. An anonymous survey asking "What AI tools do you use for work?" will surface most of it. Combine with browser extension audits and expense report reviews.
-
Classify your data. Not all data carries the same risk. Customer PII and financial records need strict controls. Generic meeting notes and brainstorm outputs are lower risk. Give your team a simple framework for what can and cannot go into external AI tools.
-
Provide sanctioned alternatives that are actually good. This is where most organizations fail. They ban the tools people love and replace them with nothing. The alternative needs to be as easy to use as the unsanctioned options, work with the data sources your team relies on, and let non-technical members build and automate without waiting for engineering.
Vybe is built for exactly this. Instead of employees cobbling together AI tools on their own, Vybe gives your team a single platform where they can build apps, automate workflows, and work with AI, all within a governed environment. It connects to 3,000+ integrations, so the tools your team already uses plug right in. And because everything runs through the platform, IT has visibility into what gets built and what data it touches.
Explore our templates library for ready-to-use starting points, or see how teams like CO2 AI handle AI governance in practice.
Get ahead of it
Shadow AI is a symptom, not a disease. The disease is a gap between what your team needs and what you have approved for them. Try Vybe free and close that gap before the next data exposure incident does it for you.
