EU AI Act Compliance Guide for RCAN Robots¶

Deadline: August 2, 2026 (General Purpose AI provisions in force)

Applicability¶

Robots using RCAN that incorporate LLM/AI capabilities fall under EU AI Act as:

High-risk AI systems (Annex III, §3 — safety components of machinery)
General Purpose AI (if using foundation models like Claude/GPT)

Required Compliance Steps¶

Article 13 — Transparency (MessageType 18)¶

Robots must be capable of disclosing:

That they are an AI system
Their operator identity
Their capabilities and limitations
Contact information for complaints

Implementation: Use RCAN MessageType.TRANSPARENCY (18) when:

A robot enters a human-occupied space
A human directly interacts with the robot
Requested via voice command or button

from rcan.safety import make_transparency_message

msg = make_transparency_message(
    source_ruri="rcan://rcan.dev/acme/arm/v1/unit-001",
    target_ruri="rcan://local/human-display",
    operator="Acme Robotics",
    capabilities=["navigation", "manipulation", "speech"],
    model_family="claude-sonnet",
    limitations=["cannot lift > 5 kg", "outdoor use only in dry conditions"],
    contact="safety@acme-robotics.example",
    p66_conformance_pct=87.5,
    audit_enabled=True,
)

Article 17 — Quality Management System¶

Deployers must maintain:

Risk assessment documentation (ISO 12100 recommended)
Test and validation records
Incident log
Human oversight procedures

RCAN Implementation: audit trail from GET /api/safety/manifest + daily audit logs

Article 9 — Risk Management¶

Apply ISO 12100 risk assessment process:

Hazard identification (robots in human spaces)
Risk estimation (likelihood × severity)
Risk reduction (Protocol 66 + RCAN safety invariants)
Residual risk documentation

Timeline¶

March 2026: MessageType 18 spec (this PR)
Q2 2026: Reference implementation in OpenCastor
August 2, 2026: Full compliance deadline