What Is Shadow AI? The Risks, and How to Bring It Inside

Shadow AI is staff using AI tools the business has not approved, usually personal accounts holding company information. It is extremely common, because people find AI useful and the official option is missing or weaker. The fix is not a ban. Instead, give people a better sanctioned option, with the business's own knowledge built in.

This is general information, not legal advice.

What is shadow AI?

Shadow AI is the use of AI tools at work without the business's approval or oversight. Common examples include pasting customer details into a personal chatbot account, using an unapproved transcription or summarising tool, and switching on AI features inside software nobody has reviewed. It is the AI version of shadow IT.

Most of it is mundane. A salesperson pastes call notes into a personal assistant to draft a follow-up. A marketer uploads a customer list to get segment ideas. Meanwhile, a manager drops a draft contract into a free tool to summarise it. None of them is trying to cause harm. They are trying to get through the week.

The problem is where the information goes and what happens to it. Data in a personal account sits outside the business's control. In addition, whatever that assistant learns about the business belongs to the employee's account, not the company.

How common is shadow AI?

Very common. Surveys consistently find that far more staff use AI than their employers have officially provided. MIT NANDA's 2025 study found workers at over 90% of companies used personal AI tools, while only 40% of companies had bought an official subscription. Microsoft and LinkedIn found 78% of AI users bring their own tools.

According to Microsoft and LinkedIn's 2024 Work Trend Index, 75% of knowledge workers use AI at work, and 78% of those users bring their own tools rather than ones their employer provides. The MIT figure is a share of companies, not of workers. Even so, both point the same way: unofficial use is the norm.

It also persists after an official rollout. According to BCG's 2025 AI at Work survey, when employees lack the AI tools they need, more than half say they will find alternatives and use them anyway. Buying licences does not end shadow AI if the licensed tool is less useful than the one people already have.

What are the risks of shadow AI?

The main risks are data leaving the business's control, breaches that cost more, privacy obligations being breached, inconsistent work, and knowledge that leaves with the employee. Each grows with the sensitivity of the information people paste in, such as customer, employee and financial data.

The security cost is measurable. According to IBM's 2025 Cost of a Data Breach report, one in five organisations studied reported a breach due to shadow AI. Organisations with high levels of shadow AI saw breach costs an average of USD 670,000 higher than those with little or none. The same report found that 63% of breached organisations had no AI governance policy, or were still developing one.

The other risks are quieter:

• Inconsistency. Each person's AI holds a different version of the business, so answers to customers vary.
• Lost knowledge. Context built up in a personal account leaves when the person does. The guide to offboarding in the AI era covers this in detail.
• No audit trail. Nobody can see what was shared, with which tool, or what came back.

What does Australian privacy guidance say about AI tools?

The Office of the Australian Information Commissioner says the Privacy Act applies to all uses of AI involving personal information. As a matter of best practice, it recommends that organisations do not enter personal information, particularly sensitive information, into publicly available generative AI tools, because of the significant and complex privacy risks.

The OAIC published this guidance in October 2024 and updated it in January 2025. For a business with shadow AI, the implication is direct. Staff pasting customer or employee details into personal chatbot accounts is exactly the practice the regulator advises against.

The guidance also covers what AI produces. Personal information that AI generates or infers about an identifiable person, including inaccurate content, is subject to the Australian Privacy Principles. That matters when AI drafts anything about a real customer or employee.

Why do employees use shadow AI?

Because it helps them and the official alternative is missing, slow or worse. People rarely use shadow AI to break rules. They use it because the approved tools do not exist yet, lack the capability they need, or know nothing more about the business than a free chatbot does.

That last reason is underrated. If the official assistant is just a company-badged version of the same general tool, it offers staff little that their personal account does not. The sanctioned option wins only when it is genuinely better, and the most reliable way to make it better is to give it the business's own knowledge.

This is why shadow AI is best read as a demand signal. It shows exactly which jobs people want help with. Those are the first workflows worth building properly.

Should you ban shadow AI?

Banning it rarely works and often makes it worse. People who find AI useful keep using it, just less visibly, and the business loses what little sight it had. A better approach is to provide an approved option that is as good or better, set clear rules on data, and make the safe path the easy one.

Bans also carry a cost. Teams that cannot use AI fall behind teams that can, and the most capable staff tend to be the ones who leave. The aim is not to stop AI use. It is to move it into accounts the business controls.

How do you bring shadow AI inside?

Find out what people use and why, provide business accounts for an approved assistant, write a short policy on what data goes where, and give the official tool the business's own knowledge so it beats personal accounts. Then fold the useful context staff have built into the shared system.

A practical sequence:

1. Ask, do not audit first. A no-blame survey of which tools people use, for what, finds more than network logs do.
2. Provide business accounts. Move staff onto an approved assistant with business data controls.
3. Write the rules on one page. Say which data may go into which tool, and what never goes anywhere.
4. Make the official tool better. Give it product facts, customer language and methods through a shared company brain.
5. Harvest what people built. Fold useful prompts and context from personal accounts into the shared system, so it becomes the company's.

Skein helps leadership teams set the rules and the rollout, then builds the shared knowledge that makes the official tool worth using. That suits a business where staff already use AI on their own. A business with strict regulatory controls on all data may need its compliance team to set the rules first.

Frequently asked questions

What is shadow AI?

Shadow AI is the use of AI tools at work without the organisation's approval or oversight. Typical examples are staff pasting company or customer information into personal chatbot accounts, using unapproved transcription or summarising tools, and switching on AI features in software that nobody has reviewed for security or privacy.

Why is shadow AI a risk?

It moves company data outside the business's control, raises breach costs and can breach privacy obligations. IBM found high levels of shadow AI added an average of USD 670,000 to breach costs in 2025. It also causes inconsistent work, and the knowledge staff build up leaves with them.

Is it legal for employees to use ChatGPT at work in Australia?

Using AI tools at work is generally lawful, but the Privacy Act applies to any use involving personal information. The OAIC recommends, as best practice, that organisations do not enter personal or sensitive information into publicly available generative AI tools. Businesses should set clear rules and use business accounts.

How do you detect shadow AI?

Start by asking staff, through a no-blame survey of which tools they use and why. That usually reveals more than technical monitoring. IT teams can add a review of software and browser extensions, network traffic to AI services, and AI features switched on inside existing applications.

How do you stop shadow AI?

Make the approved option better than the unofficial one. Provide business accounts for an approved assistant, write a one-page policy on what data may go where, train staff on their own tasks, and give the official tool the business's knowledge so it outperforms personal accounts.

Related reading