High-risk dates reflect the Digital Omnibus amendments adopted in June 2026.
Coverage Summary
Article 6 - Classification of High-Risk AI Systems
Requirement: Organizations must classify AI systems based on their risk level per Annex III categories. MeshAI coverage: The Risk Classification feature lets you assign risk levels (minimal, limited, high, unacceptable) to each agent. AI-assisted suggestions analyze agent metadata against Annex III categories.Article 12 - Record-Keeping
Requirement: High-risk AI systems must have automatic logging capabilities that record events throughout the system’s lifecycle. MeshAI coverage: The Audit Trail captures core governance actions as immutable events - agent registration/updates/deletion, kill-switch blocks, risk classification, policy changes, and anomaly acknowledgement/resolution. Events are timestamped, attributed to an actor, and exportable in CSV/JSON. (Some actions - approvals, incidents, FRIA, quarantine, ABAC, lifecycle - do not yet emit their own audit event; see each feature’s page.)Article 13 - Transparency and Information to Deployers
Requirement: High-risk AI systems must be designed to be sufficiently transparent to enable deployers to interpret outputs and use them appropriately. MeshAI coverage: Transparency Cards are auto-generated for each agent, documenting purpose, capabilities, limitations, model provider, and risk classification. Access viaGET /agents/{id}/transparency-card.
Article 14 - Human Oversight
Requirement: High-risk AI systems must be designed to allow effective human oversight during use. MeshAI coverage: HITL Approvals - via therequire_approval policy type - inline-enforce human review before agents can execute certain actions. require_human_review is a periodic-review flag: it is recorded as evidence, not enforced inline, so it documents a review cadence rather than blocking an action in real time.
Article 26 - Obligations of Deployers
Requirement: Deployers must implement appropriate technical and organizational measures, monitor AI system operation, and keep logs. MeshAI coverage: The Agent Registry provides a complete inventory of all deployed AI agents. Governance policies enforce organizational rules. Real-time monitoring detects anomalies. All actions are logged in the audit trail.Article 27 - Fundamental Rights Impact Assessment
Requirement: Deployers of high-risk AI must conduct a fundamental rights impact assessment (FRIA) before deployment. MeshAI coverage: The FRIA feature provides structured templates covering all six required assessment areas (a–f). Each FRIA is immutable once created - reassessing an agent creates a new, incrementing version rather than editing the prior one.Article 50 - Transparency Obligations for Certain AI Systems
Requirement: Providers of AI systems that interact with natural persons must ensure users are informed they are interacting with AI. MeshAI coverage: Article 50 Transparency Evidence records application-reported AI disclosure and content-marking signals from OpenTelemetry spans. Built-in, observe-only policies evaluate whether the signals are present without treating missing telemetry as proof of non-compliance. Transparency Cards document agent purpose and interaction patterns, and thelimited risk classification flags agents that may require transparency disclosures.