Shadow AI: Use AI at Work Without Leaking Data
For employees and managers: what shadow AI is, what the research really says about the risk, and 9 concrete steps to keep company and customer data safe while still using AI.
Shadow AI means using AI tools at work without your company's approval. A typical case: an employee opens ChatGPT on a personal account and pastes in company emails, financial reports or customer data, and nobody in IT knows. Once that text leaves the company's systems, the company no longer controls where it is stored or who can see it.
The fix is not to stop using AI. It is to use approved tools, keep sensitive data out of prompts, and give people a safe option that is as easy as the unsafe one. This guide shows how, with a checklist at the end.
What the research actually says
You may have seen the claim that "40% of data leaks are caused by shadow AI". That is not what the original source says. The 40% comes from a Gartner prediction about the future, not a share of today's leaks. Here are the verified numbers and exactly what each one measures.
Gartner, November 2025: "by 2030 more than 40% of enterprises will experience security or compliance incidents linked to unauthorized shadow AI." It is a forecast about organizations, not a measurement of leaks.
Gartner survey of 302 cybersecurity leaders, March to May 2025: 69% of organizations suspect or have evidence that employees are using prohibited public generative AI tools.
IBM Cost of a Data Breach 2025 (600 breached organizations): one in five organizations reported a breach due to shadow AI, and only 37% had policies to manage AI or detect shadow AI.
IBM 2025: organizations with high levels of shadow AI paid on average $670,000 more per breach than those with low or no shadow AI. Shadow AI incidents also exposed more personal data (65% vs 53% average) and intellectual property (40% vs 33%).
Microsoft and LinkedIn Work Trend Index 2024 (31,000 people, 31 countries): 75% of knowledge workers use AI at work, and 78% of those AI users bring their own AI tools to work.
IBM Cost of a Data Breach 2026 (602 breached organizations): only 40% of organizations reported using access controls on AI models and data. The global average cost of a breach reached $4.99 million.
A real case: in May 2023 Bloomberg reported that Samsung banned staff from using generative AI tools like ChatGPT after employees uploaded sensitive internal code to the platform. The concern was that data sent to these services is stored on external servers.
If you are an employee: 4 steps
- Do not use a personal AI account for company work. Ask your company for an approved work account instead. A work account is managed by your company, under its own data settings. For example, OpenAI states that by default it does not use content from ChatGPT Business, Enterprise or Edu workspaces to train its models.
- Run the "front page test" before you paste. Ask yourself: if this text were made public tomorrow, would it cause a problem? If yes, do not paste it. That rules out customer emails and names, ID numbers, financial figures, contracts, passwords, source code and internal plans.
- Check a tool is approved before you install or sign up. Ask your IT team: is this tool approved, and has it had a security review? If nobody can give a clear answer, do not use it for work. Browser extensions and "AI note takers" count too.
- Do not work around your company's security tools. If your work laptop blocks an upload or a website, that is usually a data loss prevention (DLP) control doing its job. Do not switch to your phone or personal laptop to get around it; ask IT for an approved route.
Need help with a sensitive task? Remove the details first
If an approved tool is allowed for the task, you can often get the help you need without the real data. Replace names, numbers and identifiers with placeholders, then ask:
Rewrite this email to a customer so it is clear and polite.
Keep the placeholders exactly as they are.
Dear [CUSTOMER NAME],
Your invoice [INVOICE NUMBER] for [AMOUNT] is overdue by [DAYS] days...Using ChatGPT on a personal account for your own tasks? On Free, Plus and Pro, data sharing for training is on by default. To turn it off on the web: account menu, then Settings, then Data controls, then switch off Improve the model for everyone. On the phone: open the sidebar, tap your profile icon, then Data controls. This stops training on your chats, but it does not make a personal account approved for company data.
If you are a manager or in IT: 5 steps
Gartner's advice to CIOs is to "define clear enterprise-wide policies for AI tool usage, conduct regular audits for shadow AI activity and incorporate GenAI risk evaluation into their SaaS assessment processes." In practice:
- Write a short AI acceptable use policy. Sort tools into three groups: approved, limited (allowed for some tasks or data types only) and not allowed. Say which data must never go into any AI tool. Share it with every employee and keep it to one page.
- Give people an approved alternative. Banning AI completely rarely works: people find another way because AI helps them finish work faster. Provide an approved tool on a business or enterprise plan with admin controls, so the safe option is also the easy one.
- Audit for shadow AI regularly. Use network monitoring and SaaS discovery tools to see which AI apps people actually use. Start with teams that handle sensitive data, such as finance, HR and legal. A monthly review is a reasonable starting rhythm.
- Add AI to your vendor reviews. Before approving any new software, ask whether it includes AI features, where it sends data, whether that data is used for training, and where it is stored.
- Train your team continuously. One session is not enough. Run a short quarterly workshop on the risks of shadow AI and on how to use the approved tools well, and make it part of team culture rather than a one-time warning.
A policy alone is not enough. IBM's 2025 report found that among organizations with AI governance policies, only 34% perform regular audits for unsanctioned AI. Write the policy, then check it is followed.
In the UAE: what the data protection law expects
The UAE's Personal Data Protection Law is Federal Decree by Law No. 45 of 2021, in force since 2 January 2022. Four points matter for AI at work:
Security is required
Article 20: companies that process personal data must apply appropriate technical and organisational security measures, such as encryption and pseudonymisation.
Transfers abroad are limited
Articles 22 and 23: personal data may be transferred outside the UAE only in the cases the law lists. An unapproved AI tool may process data on servers you know nothing about.
Breaches must be reported
Article 9: when a breach affects personal data, the company must notify the UAE Data Office and, where privacy is affected, the people whose data it is.
Check which rules apply
Article 2: the law does not cover government entities, some health and banking data with their own laws, or free zone companies with their own data protection laws. Those follow their own rules.
This is a summary, not legal advice. The Arabic text of the law prevails over the English translation. Ask your compliance or legal team which rules apply to your organization.
Checklist: start today
Sources: Gartner press release "Gartner Identifies Critical GenAI Blind Spots That CIOs Must Urgently Address" (19 November 2025); IBM Cost of a Data Breach Report 2025 newsroom release (30 July 2025) and 2026 report pages (29 July 2026); Microsoft and LinkedIn 2024 Work Trend Index (8 May 2024); Bloomberg, "Samsung Bans Staff's AI Use After Spotting ChatGPT Data Leak" (2 May 2023); OpenAI Help Center, "Data controls in ChatGPT"; UAE Federal Decree by Law No. 45 of 2021 on the Protection of Personal Data (uaelegislation.gov.ae). Checked November 2026.
Want your team to use this at work?
I'm Abdulaziz Alshamsi, a certified AI trainer and consultant in Abu Dhabi. I train government and company teams to use AI in their daily work. Message me to discuss a session for your team.