The EU AI Act (Regulation 2024/1689) establishes requirements for AI systems operating in the EU. Key obligations take effect August 2, 2026.
High-risk AI systems shall be designed with logging capabilities that record events relevant to the functioning of the AI system.
| Requirement | How Capsule Addresses It |
|---|---|
| Automatic logging of events | Every AI action produces a Capsule (the axiom: for all actions, there exists a Capsule) |
| Traceability of results | Execution section records tool calls; Outcome section records results and side effects |
| Monitoring of operation | Chain provides temporal ordering; session_id groups related interactions |
| Identification of risks | Reasoning section captures risk assessment in ReasoningOption.risks |
High-risk AI systems shall be designed to ensure their operation is sufficiently transparent.
| Requirement | How Capsule Addresses It |
|---|---|
| Understandable output | Outcome section includes human-readable summary field |
| Explanation of decisions | Reasoning section captures analysis, options, and rationale before execution |
| Information for deployers | Capsules are queryable via storage backends; all fields are machine-readable |
High-risk AI systems shall be designed to be effectively overseen by natural persons.
| Requirement | How Capsule Addresses It |
|---|---|
| Human-in-the-loop capability | Authority section supports human_approved type with approver identity |
| Ability to intervene | Kill switch Capsules (CapsuleType.KILL) record intervention events |
| Override capability | Authority section's escalation_reason documents why human override occurred |