1. Purpose
This plan defines how TFG prepares for, detects, contains, eradicates, recovers from and learns from information-security, cyber, privacy and AI-related incidents. It supports UK GDPR, the Data Protection Act 2018 and ISO 27001-aligned incident management.
2. Incident types
Unauthorised access, malware, ransomware, data loss, account compromise or service outage.
Client or personal information disclosed through a prompt, output, connector, log or knowledge source.
Prompt injection, jailbreak, unsafe content, unauthorised tool execution or agent action.
Excessive permissions, misconfigured retrieval, cross-client exposure or unintended source access.
Model/provider incident, unexpected retention, sub-processor change or service compromise.
Incorrect high-impact output acted upon without required human approval.
3. Severity
Severity
Definition
Examples
P1 Critical
Severe legal, client, safety, confidentiality, integrity or availability impact.
Confirmed sensitive-data exposure; material cross-client disclosure; ransomware; uncontrolled high-impact agent action.
P2 Major
Serious incident with potential escalation or material service/client impact.
Unauthorised access; high-risk prompt leakage; agent misconfiguration; malware detected early.
P3 Minor
Limited incident requiring remediation.
Low-impact inappropriate output; contained misconfiguration; suspicious activity.
P4 Informational
No confirmed impact; monitor or record.
Threat intelligence; blocked policy violation; unsuccessful attack.
4. Roles
Role
Responsibility
SOC / Simplify IT
Monitoring, investigation, containment, recovery and evidence preservation.
Operations Director / ISMS Lead
Incident coordination, severity, personal-data assessment, client/regulatory decision support and incident log.
AI / Product Owner
Disable affected agent or connector; preserve prompts, configuration, sources and evaluations; support root-cause analysis.
Security / Privacy / Legal
Assess contractual, privacy and notification obligations.
Board
Risk acceptance, strategic decisions and business continuity.
All personnel
Report suspected incidents immediately and preserve evidence.
5. Response process
Prepare
Maintain contacts, logging, backups, access controls, training, playbooks and tested shutdown routes.
Detect and triage
Log the report; identify systems, data, users, prompts, sources, connectors and actions; classify severity and preserve evidence.
Contain
Disable compromised accounts, agents, connectors or tools; revoke tokens; isolate systems; suspend data flows; block malicious activity.
Eradicate
Remove malicious content or code, correct permissions and configuration, rotate secrets, patch vulnerabilities and remove unsafe grounding data.
Recover
Restore in a controlled manner, retest permissions and safety controls, increase monitoring and communicate with authorised stakeholders.
Review
Complete root-cause analysis, lessons learned, client/regulatory review, corrective actions and policy/model updates.
6. Notification
The Operations Director and Privacy/Legal support assess whether the incident is a personal-data breach and whether notification duties apply. Where the statutory threshold is met, the ICO notification deadline is 72 hours after awareness. Contractual client notification requirements may be different and must be checked.
7. AI evidence checklist
Prompt and response; timestamp; user and agent identity.
System instructions, model/version, tools and connectors.
Sources retrieved and permission context.
Actions attempted or completed.
Relevant logs, screenshots, correlation IDs and configuration versions.