Analysis Cybersecurity AI Compliance

Shadow AI: the data leak hiding in your organisation's everyday tools

Across many organisations, employees paste customer data, financial records and confidential plans into AI tools their employer has never assessed, often through personal accounts. Most IT and security teams cannot see it happening. The exposure already exists, and it touches data protection, confidentiality and audit evidence at the same time.

10 September 2026
8 min read
Updated 10 November 2026
Key takeaways
  • In a Censuswide survey of 2,003 UK employees for Microsoft (October 2025), 71% said they had used AI tools their employer had not approved, and 51% said they still do so every week
  • Personal data pasted into an unapproved AI tool can breach UK GDPR principles even when the employee meant well. Your organisation, as controller, has to show the processing was lawful, secure and covered by the right terms
  • On personal consumer accounts, some providers use conversations to improve their models unless the user opts out; business and enterprise offerings generally do not by default. Check the current terms for each tool
  • ISO 27001 expects you to know which information assets and cloud services you use. AI tools your IT team has not identified are a gap in that inventory and in your risk assessment
  • Blocking access without offering an approved alternative tends to push use onto personal devices, where your controls cannot see it

What shadow AI looks like in practice

Shadow AI is rarely a single tool or a deliberate breach of policy. It builds up from many small decisions that each seem reasonable. A sales manager drafts a client proposal in ChatGPT because it is faster. A developer asks an AI assistant to review a block of code. An HR professional pastes a job description and a candidate's CV into a free AI tool to help score responses. An analyst summarises a board report with a consumer AI service because the company has not provided an alternative.

None of these people are acting maliciously. They found tools that make their work easier. The problem is the gap between what they intend and the risk your organisation takes on as a result.

Shadow AI is the AI-era version of shadow IT: technology used without IT oversight. With shadow IT, the usual concern is unsupported systems, unpatched software and data stored in the wrong place. With shadow AI, staff type or paste your information into external services, and once it has been sent you cannot reliably get it back.

Microsoft's research into UK workplace AI use, carried out by Censuswide among 2,003 UK employees in October 2025, found that 71% had used AI tools their employer had not approved and 51% continue to do so every week. Verizon's 2026 Data Breach Investigations Report points the same way from a different data set: among employees regularly using AI on corporate devices, 67% did so through non-corporate accounts. The two studies measure different things, so read them as separate signals.

71%
of UK employees surveyed had used AI tools their employer had not approved (Microsoft, 2025)
67%
of regular AI users on corporate devices used non-corporate accounts (Verizon DBIR 2026)
4%
of worldwide annual turnover, or £17.5m if higher: the maximum UK GDPR fine for the most serious infringements

The three risks that matter most

Most discussion of shadow AI focuses on data security. Three separate kinds of risk come together when staff use an unapproved AI tool at work, and the compliance exposure goes well beyond a stolen password.

1. UK GDPR and data protection

UK GDPR Article 5 requires personal data to be processed lawfully, fairly and transparently, for specified purposes and with appropriate security, and Article 6 requires a lawful basis. When a staff member pastes a customer record, patient information or HR data into an AI service through a personal account, your organisation has usually not assessed the provider, told the individuals concerned, or agreed any terms on retention, security or reuse.

The provider's role depends on the account and the terms. Where an AI provider processes personal data on your behalf, Article 28 requires a written contract with specific processor terms. On a personal consumer account the provider has no contract with you at all: it deals with the employee under its consumer terms and may decide for itself how conversations are used, which can make it a controller of that data in its own right. Either way, nothing you have agreed covers the disclosure.

Responsibility does not pass to the employee because they acted on their own initiative. Your organisation is accountable for the personal data it controls (Article 5(2)) and must have appropriate technical and organisational measures in place (Articles 24 and 32). Whether a specific disclosure leads to ICO action or a compensation claim under Article 82 depends on the data, the harm and what you did about it.

International transfer rules add another layer. Many AI services process data outside the UK. Under UK GDPR Chapter V, a restricted transfer needs UK adequacy regulations, appropriate safeguards such as the International Data Transfer Agreement (IDTA) or the UK Addendum to the EU standard contractual clauses, or one of the limited exceptions. A personal account used without your knowledge has none of these in place.

2. Confidentiality and intellectual property

Beyond personal data, staff regularly paste commercially sensitive material into AI tools: source code, acquisition plans, client contracts, pricing and competitive research. On personal accounts, some providers use conversations to improve their models by default, with an opt-out in the settings. OpenAI, for example, explains that content from personal ChatGPT accounts may be used to improve its models unless the user turns this off, while data from ChatGPT Business, Enterprise, Edu and the API is not used for training by default. Terms differ between providers and change over time, so check them for each tool.

Even where a provider does not train on inputs, the information has left your control. It may be retained, reviewed for abuse monitoring, exposed in a breach at the provider, requested by an authority in another jurisdiction, or mixed in with the employee's personal use of the same account. Client contracts and NDAs may also forbid sharing it with any third party.

Your staff are not trying to leak data. They are reaching for tools that work, and a personal account on those tools carries no data governance commitments to your organisation.

3. Inaccurate outputs used without verification

AI tools produce confident-sounding outputs that are sometimes wrong. When staff use those outputs in customer-facing documents, compliance filings or internal decisions without checking them, your organisation carries the accuracy risk. This often runs alongside the data exposure: the person pasting sensitive data in is frequently the same person relying on the answer without cross-checking it.

In regulated sectors, incorrect AI-generated content in medical, legal or financial advice can create liability far larger than the time it saved.

The legal exposure and how to get control

Two companion articles cover what to do about it.

Shadow AI and UK GDPR explains why a personal AI account creates exposure under Article 28, when an unapproved prompt becomes a reportable breach, what the NIS Regulations and ISO 27001 expect, and the steps to take when data has already gone into an unapproved tool.

Finding and controlling shadow AI covers the places it hides beyond ChatGPT, including AI features inside approved SaaS tools, and the seven components of a response your staff will follow.

Where to start

Shadow AI is probably already happening in your organisation. Acting now costs less than acting after a data incident or a regulatory inquiry.

A useful first step is a short, anonymous staff survey asking which AI tools people use and for what. The answers often differ from what IT expects, and they show you where to focus. Policy and tool decisions based on real evidence tend to hold up better than ones based on assumptions.

Three actions are worth prioritising:

  • Publish an AI acceptable use policy, even a short interim version, before your next staff review cycle. Staff then have a clear rule to follow, and you have something to point to if an incident occurs.
  • Review your processor register and identify every AI service that processes personal data on your behalf. Confirm each one is covered by a written Article 28 contract and, where data leaves the UK, a valid transfer mechanism.
  • Audit your existing SaaS tools for AI features. Many have added capabilities since your original vendor assessment. Update your records and, where needed, your contracts.

Shadow AI is a gap between how your staff work and governance designed before these tools existed. Close it with policies people can follow and approved tools that give them the productivity they have already found on their own.


How Cyvra helps

Cyvra combines security and AI expertise to help organisations understand where shadow AI is happening, assess the GDPR and confidentiality exposure, and build a response your staff will follow, by giving them better tools rather than a blanket prohibition.

  • Shadow AI audit: identify which AI tools your staff are using, including AI features embedded in SaaS platforms your IT team approved for other purposes
  • UK GDPR processor review: identify which AI services process personal data on your behalf without an appropriate contract, and assess the exposure
  • AI acceptable use policy: draft a policy that gives your staff clear guidance on approved tools and permitted data types, distinguishing between use cases rather than issuing a blanket prohibition
  • Approved AI tooling: recommend enterprise-grade alternatives that meet your data governance requirements, so staff have an approved route to the productivity they are looking for
  • Staff awareness training: run targeted sessions that explain data classification in the context of AI tools, not generic annual compliance exercises
  • DLP configuration: configure data loss prevention controls to detect and block sensitive data transfers to unapproved AI services

Contact our cybersecurity or AI practice to start with a shadow AI audit.

Shadow AI tools that process personal data at scale, or that are used in automated decision-making, may trigger Data Protection Impact Assessment obligations. Our guide to AI DPIA requirements covers when an assessment is required and how to conduct it.

For organisations deploying AI that can take actions autonomously, the security considerations extend beyond data exposure. Our guide to agentic AI security risks covers governance and control requirements for AI systems with reduced human oversight.

Frequently asked questions

What is shadow AI?

Shadow AI is the use of AI tools by your staff without the knowledge or approval of your IT or security team. This includes personal ChatGPT accounts, consumer-tier Claude or Gemini access, AI features embedded in third-party SaaS tools, and browser extensions with AI capabilities. Unlike shadow IT, where the concern is unsupported systems, shadow AI means your staff are actively inputting sensitive content into external services with no data governance controls in place.

Do free AI tools use your inputs to train their models?

Many consumer AI services can use conversations to improve their models, though policies vary by provider and change over time. OpenAI, for example, may use content from personal ChatGPT accounts for model improvement unless the user opts out in the data controls, while data from ChatGPT Business, Enterprise, Edu and the API is not used for training by default. Check the current terms for each tool and direct staff to approved business accounts.

Ryland Deakin
About the author
Lead Consultant, Cyvra · CISM · CompTIA Security+ · MCP

Ryland has delivered cybersecurity, compliance, and IT management programmes for regulated organisations across the UK and the Netherlands for over 20 years. View full profile

Talk to Cyvra

Map your shadow AI exposure

We help you identify which AI tools your staff are using, build a policy they will follow, and deploy enterprise tools that give them a compliant alternative.

Disclaimer: This article is for general informational purposes only and does not constitute legal, regulatory, or professional advice. Cyvra makes no warranty as to the accuracy or completeness of this content. Readers should seek independent advice appropriate to their specific circumstances. Cyvra accepts no liability for any loss arising from reliance on this content.