TL;DR

The National Cyber Security Centre has weighed in on staff using AI tools their employer never approved, and its advice is not what a nervous IT director might hope for. Prohibition is not the recommendation. The agency’s position is that the practice will not go away, that policy has lagged behind what employees need, and that the workable goal is reducing risk rather than pretending it can be removed.

What the agency is describing

Shadow AI is a variant of shadow IT: tools running outside the systems and processes an organisation has sanctioned. The NCSC cites research finding 71% of employees using AI tools their employer has not approved — a figure high enough that treating it as a discipline problem misreads the situation.

The stated cause is a gap rather than defiance. Where security policy cannot meet a business need, staff adopt services before anyone has assessed them, and cheap, capable tools make that easier every month.

The three risks named

Sensitive information leaving the organisation is the obvious one. Feeding company or customer data into an unapproved service raises exposure to breaches, loss of intellectual property and regulatory failure.

Loss of visibility is subtler and, for governance purposes, worse. Information handed to a consumer AI service can be kept, held indefinitely, or fed back into training the product, outside whatever controls the organisation thinks it operates — unless specific privacy settings are switched on.

The third is the one that has changed most recently. AI agents are complex software with exploitable vulnerabilities, and an attacker who compromises one inherits everything it can legitimately reach: its data, its services, its privileges. The NCSC expects attackers to favour agents with looser guardrails precisely because they make useful entry points into the wider corporate estate.

Looking forward

The recommended response is cultural before it is technical. Organisations with open communication about security are less likely to see staff route around it, and knowing why people reach for a given tool is what makes it possible to offer a secure alternative rather than a refusal.

This lands days after our coverage of research finding most UK bosses already assume shadow AI is running inside their firms. The NCSC has now supplied the official framing for a problem those leaders had already identified — which is useful, because “you cannot manage what you do not know” is easier to act on when it carries a government logo.