Does using Copilot require any extra CE evidence?

Not specifically – Copilot inherits tenant controls. Make sure your SharePoint and OneDrive permissions are tight, because Copilot will surface what the user can already access.

Is ChatGPT Enterprise in scope?

Yes. It is a SaaS app holding organisational data and authenticating users. List it in your scope, enforce SSO + MFA, and configure data controls.

Can I ban AI to simplify CE?

You can, but staff will use it anyway. A managed AI policy with a sanctioned tool is far stronger than a blanket ban.

Do I need to mention AI in my scope statement?

Yes if the AI tool is an in-scope cloud service. Embedded AI like Copilot can be covered by the M365 entry; standalone AI SaaS gets its own line.

Does the free CE insurance cover AI-related incidents?

The free £25,000 IASME cyber policy responds to cyber incidents regardless of cause, subject to policy terms. AI misuse leading to a data breach would generally trigger the cyber response, but always check the wording.

NCSC AI Guidance and Cyber Essentials Scope: Copilot, ChatGPT, Shadow AI

The three categories of AI tool

Embedded AI – Microsoft 365 Copilot, Google Workspace Gemini. Inherits tenant identity and data controls.
Enterprise SaaS AI – ChatGPT Enterprise, Claude Teams, Glean. A SaaS app in scope like any other.
Shadow AI – staff using consumer accounts (free ChatGPT, free Claude) with work data. Out of your control by default.

Embedded AI (Copilot, Gemini)

Because Copilot runs inside your Microsoft 365 tenant, it inherits Conditional Access, MFA and DLP. The CE/Danzell scope question is the same as for the underlying tenant. Make sure SharePoint and OneDrive permissions are tight – Copilot will surface whatever the user can already see, and over-permissive sites become an instant data leak.

Enterprise SaaS AI (ChatGPT Enterprise et al.)

Treat exactly like any other SaaS app:

SSO from your identity provider (Entra ID, Okta, Google).
MFA enforced via the IdP.
Admin console MFA – see Danzell MFA guide.
Data retention and training-opt-out configured at tenant level.
Listed in your scope as an in-scope cloud service (SaaS).

Shadow AI: the real risk

A staff member pasting client data into free ChatGPT is the most common AI incident UK assessors flag. Practical controls:

Acceptable use policy that names AI tools explicitly.
Provide a sanctioned AI tool so staff have a safe alternative.
Browser-side controls or DLP for high-risk roles.
Identify and block consumer AI domains where the risk justifies it.

NCSC guidance in one paragraph

NCSC's position is consistent with its wider guidance: understand where your data goes, control identity and access, monitor for misuse, and treat AI providers like any other third-party processor. Nothing in Cyber Essentials requires you to avoid AI – it requires you to bring AI tools into scope properly.