§24 Instructions for Use¶
Version: v3.0
Defines the rcan-ifu-v1 schema — a structured Instructions for Use document covering all eight fields required by EU AI Act Art. 13(3) for high-risk AI systems.
Overview¶
EU AI Act Art. 13 requires providers of high-risk AI systems to accompany each system with clear and adequate instructions for use. Art. 13(3) specifies eight mandatory information categories. RCAN §24 defines a machine-readable IFU document that can be auto-generated from the RCAN config and deployment context, then provided to deployers alongside the FRIA.
The annex_iii_basis must be one of the ten valid Annex III categories (see §22). An invalid value causes generation to fail immediately.
Art. 13(3) Field Coverage¶
| Field | Article | Content |
|---|---|---|
provider_identity |
Art. 13(3)(a) | Provider identity: RRN, robot name, contact, RCAN version, AI model. |
intended_purpose |
Art. 13(3)(b) | Intended purpose: deployment description and Annex III basis. |
capabilities_and_limitations |
Art. 13(3)(c) | What the system can and cannot do; known limitations. |
accuracy_and_performance |
Art. 13(3)(d) | Performance evidence — links to §23 benchmark results. |
human_oversight_measures |
Art. 13(3)(e) | HiTL gates, ESTOP, confidence gates, operator override. |
known_risks_and_misuse |
Art. 13(3)(f) | Foreseeable misuse scenarios and mitigations. |
expected_lifetime |
Art. 13(3)(g) | Software support lifecycle and hardware dependency. |
maintenance_requirements |
Art. 13(3)(h) | Update cadence, conformance checks, incident logging. |
Schema¶
{
"schema": "rcan-ifu-v1",
"generated_at": "2026-04-11T09:00:00.000Z",
"art13_coverage": [
"provider_identity", "intended_purpose", "capabilities_and_limitations",
"accuracy_and_performance", "human_oversight_measures",
"known_risks_and_misuse", "expected_lifetime", "maintenance_requirements"
],
"provider_identity": {
"rrn": "RRN-000000000001",
"rrn_uri": "rrn://org/robot/model/id",
"robot_name": "my-robot",
"provider_name": "Example Corp",
"provider_contact": "safety@example.com",
"rcan_version": "3.0",
"agent_provider": "anthropic",
"agent_model": "claude-sonnet-4-6"
},
"intended_purpose": {
"description": "Indoor autonomous navigation for warehouse logistics",
"annex_iii_basis": "safety_component",
"deployment_context": "High-risk AI system under EU AI Act Annex III"
},
"capabilities_and_limitations": {
"summary": "Universal robot runtime connecting LLM AI providers to physical hardware.",
"known_limitations": [
"AI provider responses subject to model confidence thresholds",
"Physical limits enforced by BoundsChecker configuration",
"HiTL authorization required for high-risk actions"
]
},
"accuracy_and_performance": {
"note": "Quantified evidence via castor safety benchmark. Safety path P95: ESTOP 100ms, BoundsCheck 5ms, ConfidenceGate 2ms, FullPipeline 50ms."
},
"human_oversight_measures": {
"hitl_gates": "HiTL authorization gates (§8) prevent autonomous high-risk actions",
"estop": "Emergency stop halts all motion; P95 latency ≤ 100ms",
"confidence_gates": "AI commands below confidence threshold are blocked",
"override": "Operators can halt the system at any time via ESTOP"
},
"known_risks_and_misuse": {
"foreseeable_misuse": [
"Deployment outside declared intended use",
"Operating beyond configured physical bounds",
"Disabling HiTL gates without risk assessment"
],
"mitigations": [
"BoundsChecker enforces hard physical limits",
"Conformance checks detect disabled safety features"
]
},
"expected_lifetime": {
"software_support": "Subject to OpenCastor release lifecycle (YYYY.MM.DD versioning)",
"hardware_dependent": true,
"note": "Deployer is responsible for post-market monitoring per Art. 72 and §25"
},
"maintenance_requirements": {
"software_updates": "Run castor upgrade for runtime updates",
"conformance_checks": "Run castor validate before each deployment",
"incident_logging": "Run castor incidents record for safety-relevant events"
}
}