Free Template · Incident Response

Shadow AI
Incident Response Plan

A complete incident response plan for shadow AI security incidents. Covers detection through recovery with severity levels, communication templates, and GDPR breach assessment guidance.

Download Template Book Demo

68%

of AI incidents involve shadow AI

72h

GDPR notification window

4 severity

levels with clear triggers

Free

to use and customise

Why Shadow AI Incidents Need a Dedicated Response Plan

Shadow AI incidents have unique characteristics that standard IT security incident response plans are not designed to handle. Unlike a traditional data breach, you often cannot recall data submitted to an AI service, and the legal implications of AI provider training terms create novel GDPR exposure that requires specialist assessment.

68%

of AI-related data incidents involve shadow AI tools

The majority of AI security incidents originate from unapproved tools used outside IT oversight, not from sanctioned AI deployments.

72h

GDPR notification window starts when you become aware

Shadow AI incidents involving personal data may trigger GDPR Article 33 notification obligations, the clock starts on awareness, not on breach occurrence.

83%

of organisations lack a shadow AI incident response process

Most IR plans predate widespread AI adoption and have no provisions for AI-specific incident types, data exposure scenarios, or AI provider terms.

6.4M

records exposed per average AI data incident

AI tools can process and potentially expose large volumes of data quickly, making the scope of a shadow AI incident larger than a typical insider data theft.

The Incident Response Plan

Click each phase to expand. Customise the highlighted placeholders and adapt severity thresholds to your organisation's risk appetite.

Shadow AI incidents are classified by the severity of data exposure and the regulatory implications. Use this matrix to determine the appropriate response track.

SEV-1 Critical

Confirmed exposure of Restricted data (PII of 100+ individuals, credentials, health data, financial account data) to an external AI service with potential training data retention. Regulatory notification likely required.

Immediate escalation to CISO + DPO. Incident Commander activated. 72-hour GDPR clock may be running.

SEV-2 High

Confirmed exposure of Confidential data (strategic plans, IP, contracts, limited PII) to unapproved AI service. No confirmed training data retention but cannot be excluded.

Security team lead notified within 2 hours. Legal/Privacy engaged. Containment initiated same business day.

SEV-3 Medium

Unapproved AI tool usage confirmed with Internal-classified data. No personal data confirmed but investigation required to verify scope.

Security analyst assigned within 4 hours. Manager of affected employee notified. Investigation initiated.

SEV-4 Low

Unapproved AI tool usage confirmed with Public or non-sensitive Internal data only. Policy violation but no data exposure risk identified.

Logged and tracked. Manager notified. Policy reminder issued. No emergency response required.

Download Full Plan

How to Implement This Response Plan

An incident response plan only works if it has been operationalised before an incident occurs. Follow these steps to go from template to live process.

Establish your shadow AI baseline before an incident occurs

Run a discovery exercise using DNS logs, proxy traffic analysis, and browser extension audits to identify which AI tools are already in use across the organisation. You cannot respond effectively to incidents involving tools you don't know about.

Define severity levels and adapt triggers to your context

Tailor the severity classification matrix to your organisation's data classification scheme and regulatory obligations. Ensure every member of the incident response team understands what triggers escalation from one severity level to the next.

Assign named owners to each incident response role

Name specific individuals (with alternates) for Incident Commander, Security Analyst, DPO/Legal Lead, Communications Lead, and Technical Remediation. Distribute the plan to all named responders and ensure it is accessible outside corporate systems in case of a major incident.

Test the plan with a realistic tabletop exercise

Run an annual tabletop exercise simulating a shadow AI incident, for example, an employee who submitted customer PII to a public LLM for summarisation. Walk through every stage of the plan and document gaps identified. Update the plan before the exercise findings are lost.

Integrate with your existing ITSM and security incident processes

Shadow AI incidents should flow through the same incident management system as other security incidents. Configure your ticketing system with a shadow AI incident category, set automated escalation rules for SEV-1 and SEV-2 incidents, and ensure the GDPR 72-hour notification clock is tracked automatically.

FAQ

Frequently Asked Questions

What is a shadow AI incident?

A shadow AI incident is any security, privacy, or compliance event arising from the use of unapproved AI tools, tools employees are using without IT or security team knowledge or authorisation. Common examples include: an employee submitting customer PII to a public AI chatbot, confidential financial data entered into an AI summarisation tool, source code uploaded to a cloud AI coding assistant, or sensitive HR data processed by a consumer AI service. Shadow AI incidents are distinct from incidents involving approved AI tools because the organisation typically has no visibility, no data processing agreement, and no ability to exercise data subject rights with the AI provider.

Does a shadow AI data exposure require GDPR notification?

Whether a shadow AI incident requires GDPR breach notification depends on the nature of the data exposed and the likelihood of harm to data subjects. If personal data was submitted to an external AI service that may have processed or stored it, particularly if the AI provider's terms allow training on user inputs, this is likely a personal data breach under GDPR Article 33. You have 72 hours from becoming aware of the breach to notify your supervisory authority if it poses a risk to individuals. Engage your DPO immediately when personal data exposure is confirmed or suspected. Do not assume no notification is required without a documented legal assessment.

How do you detect shadow AI usage in an organisation?

Detecting shadow AI requires layered technical controls. DNS and web proxy logs will show traffic to AI services such as ChatGPT, Claude, Gemini, and hundreds of AI-powered SaaS tools, most security teams are surprised by the volume when they first look. Browser extension audits reveal AI coding assistants and productivity tools installed by employees. Data loss prevention (DLP) tools with AI-specific patterns can detect sensitive data being submitted to AI endpoints. Dedicated AI security platforms like Aona provide continuous discovery and classification of all AI tool usage across the organisation, including tools accessed via personal devices on corporate networks.

What should be included in a shadow AI post-incident review?

A post-incident review for a shadow AI incident should cover: a timeline of the incident from initial use through discovery and response; root cause analysis (why was the unapproved tool used, gap in approved toolset, lack of awareness, deliberate circumvention?); impact assessment (what data was exposed, to which AI provider, for how long, and what are the ongoing risks from training data retention?); effectiveness review of detection and response actions; specific remediation actions with owners and deadlines (including technical controls to prevent recurrence); and a review of the incident response plan itself to address any gaps identified during the response. Key outputs should include updated AI governance policies, enhanced technical controls, and targeted employee awareness training.

Get started

Detect Shadow AI Before It Becomes an Incident

Aona continuously discovers unapproved AI tools across your organisation, detects sensitive data being submitted to external AI services, and alerts your security team in real time, before a shadow AI incident becomes a GDPR breach.

Book a demo Download Template

Related Resources

Aona AI Platform AI Acceptable Use Policy AI Risk Assessment Checklist All Templates