ADGM: AI-specific cyber risks enter supervisory guidance
Regulated firms should assess AI manipulation and impersonation risks in systems they develop, buy or integrate.
Official sourceRead the change, dates and action to review
Instrument: FCCP Notice 222/2026. Status: Issued practical supervisory guidance; no new rule.
Authority and jurisdiction: ADGM FSRA / FCCP; ADGM.
Publication: 10 November 2026. Effective/compliance date: No separate deadline. Existing GEN 3.5 obligations continue.
Confirmed position: The notice identifies prompt injection, data poisoning, model manipulation, synthetic identities, payment-authorisation attacks and critical third-party disruption. It recommends AI threat modelling, security testing, layered identity checks, provider oversight and recovery testing.
Impacted workflow: AI agents, onboarding, payment approvals, API permissions and outsourced technology for ADGM-regulated firms.
Interpretation and action to assess: Assess AI-specific release tests, least-privilege tool access, out-of-band verification for sensitive instructions and incident escalation. These are implementation assessments; the notice does not automatically bind a mainland technology provider.
Urgency: Before next release; immediate assessment for a live ADGM-regulated workflow.
DIWAN regulatory briefing. Applicability and implementation require assessment for the relevant entity and workflow. Last checked 12 November 2026.