Shadow AI

What is shadow AI?

Shadow AI is any AI tool that people at your company use for work without the company knowing about it or approving it. It can be a chatbot on a personal account, a browser extension that summarizes pages, a meeting note-taker someone added to calls, or an AI feature switched on inside an app you already pay for.

It is a newer form of shadow IT, the long-standing problem of software bought or used outside official channels. The difference is what happens to the data. A shadow project tracker stores your tasks. A shadow AI tool may read, keep and learn from whatever people paste into it.

Shadow AI does not require bad intent. People pick up these tools to save time. The risk comes from the terms that govern the tool, which no one at the company has read.

Common forms of shadow AI

• Personal accounts on AI chatbots used for work tasks, such as drafting emails or reviewing contracts.
• Free AI tools for writing, coding, transcription or image creation, signed up for with a work email.
• Browser extensions and plugins that send page content to an AI service.
• AI features inside approved apps that were added after you bought the app and turned on without review.
• AI connectors that link an outside assistant to your CRM, files or inbox.

What it looks like in a vendor's terms

Shadow AI is not a clause, but the terms of the tools people pick decide the risk. A typical consumer AI clause reads like this:

We may use the content you provide, including prompts, files and responses, to improve and train our models. You can change this in your settings. These terms apply to individual accounts; business plans are covered by separate terms.

• "May use the content you provide" means work data pasted into a personal account can be used by the provider.
• "You can change this in your settings" means the protection depends on each person finding and using a setting.
• "Individual accounts" means your company's business agreement, if you have one, does not cover this use.

The gap between plans can be real. OpenAI's business data page says it does not train on data from its business plans or API platform by default. That promise covers those business products, not a personal account someone opened on their own. Our page on enterprise vs consumer AI terms explains the split.

A real change we caught

Terms for popular AI tools can shift after people start using them. Venpo flagged that Lovable now trains its AI on your prompts and code by default. If a developer on your team had been using it on their own, the change would have reached your code without anyone reviewing it.

Why it matters to you

Confidential data leaves without a contract. When work data goes into a tool you have no agreement with, there is no DPA, no confidentiality clause and no limit on training beyond what the tool's public terms say.

You can break promises to your customers. Your contracts may say customer data goes only to listed subprocessors. A teammate pasting a customer's file into an unapproved AI tool can put you outside those promises.

Your security answers become inaccurate. Security questionnaires may ask which AI tools touch customer data. If your list is incomplete, your answers are too.

You miss changes. You cannot monitor terms for a tool you do not know exists. Training defaults, retention rules and data use clauses can change at any time.

Costs can creep in. Personal subscriptions charged to company cards add up and can duplicate tools you already pay for, a pattern known as SaaS sprawl.

What to check

• Ask teams which AI tools they use for work, including free tools, extensions and note-takers, and add them to your vendor inventory.
• Review expense reports and card statements for AI subscriptions.
• For each approved app, check whether new AI features were added and whether they are on by default.
• Check which AI connectors are linked to your CRM, file storage and email.
• For each AI tool in use, read its terms for training, retention and data sharing, and decide whether it needs a business plan.
• Write a short policy that names approved AI tools and says what data must never go into others.

How Venpo helps

Once a tool is on your list, Venpo monitors its terms, privacy policy and AI terms and emails you in plain English when training, retention or data use rules change. The Fine Print Scan answers, for each vendor, whether it trains AI on your data and how long it keeps it.