EU AI Act Compliance — ROBOT.md¶
Five RCAN §22-26 attestation packets — FRIA, safety-benchmark, IFU, incident-report, EU-register. Each packet is produced by a robot-md emit-* command, signed with the operator's release key, and submitted to RRF intake at /v2/<packet-type>. RRF stores the evidence in its transparency log; jurisdictional filing remains the operator's responsibility.
Conformance is not certification.
Conformance to RCAN tracks (L1–L4 protocol, Gateway Authority, HIL Runtime Safety) is self-asserted via signed bundles and independently replayable from those bundles. Conformance is not certification. Certification requires audit by a qualified third-party body, which is intentionally out-of-scope for the foundation in 2026.
Compliance evidence is not regulatory sufficiency.
Compliance packet generation and RRF submission produce evidence; they do not constitute regulatory sufficiency in any jurisdiction. Per-jurisdiction conformity assessments and notified-body engagement are the user's responsibility, in consultation with qualified counsel.
§ 01 — The 5 packets¶
Five attestation packets. One command each. Each packet covers a different EU AI Act obligation. All are signed with your robot's key and routed to the live RRF endpoint.
§22 — FRIA (Fundamental Rights Impact Assessment)¶
EU AI Act Art. 9: pre-deployment assessment of impact on fundamental rights. Automated checklist derived from ROBOT.md capabilities, safety gates, and RCAN LoA level. Filed before first deployment.
Endpoint: /v2/fria — robotregistryfoundation.org
§23 — Safety benchmark (ISO 42001 self-assessment)¶
ISO 42001 + RCAN §23: structured safety self-assessment linked to the robot's RRN. Covers hazard analysis, control measures, and residual risk. Signed envelope with audit-trail hash.
Endpoint: /v2/safety-benchmark
§24 — IFU Art. 13(3) (Instructions for Use)¶
EU AI Act Art. 13(3): machine-readable instructions for use — what the robot can and cannot do, under what conditions, with what safety constraints. Derived from ROBOT.md capabilities and body sections.
Endpoint: /v2/ifu
§25 — Incident report (Art. 72 serious incident notification)¶
EU AI Act Art. 72: structured incident report — timestamped, linked to RRN, signed. Required within 15 days of a serious incident. robot-md provides the structure; the operator provides the incident narrative.
Endpoint: /v2/incident-report
§26 — EU Register Art. 49 (Per-model EU AI Act registration)¶
EU AI Act Art. 49: registration of the AI system in the EU database. Requires an RMN (model number). Routed through RRF to the official register endpoint. robot-md handles the RCAN-signed envelope.
Endpoint: /v2/eu-register
§ 02 — Status check¶
Check filing status in one command.
Example output:
# robot-md compliance status ./ROBOT.md
# RRN: RRN-000000000003 | RMN: RMN-000000000004
✓ fria ACCEPTED 2026-04-24T11:32:01Z
✓ safety-benchmark ACCEPTED 2026-04-24T11:33:15Z
✓ ifu ACCEPTED 2026-04-24T11:34:02Z
✓ incident-report NONE no incidents filed
✓ eu-register ACCEPTED 2026-04-24T11:35:40Z
Ready for submission · 4/5 packets filed · 0 incidents
Note: pass the file path (./ROBOT.md), not a directory, to avoid issue #32 (directory-vs-file resolution bug, tracked).
§ 03 — Identity prerequisite¶
Register first. File second.
All five compliance packets require a valid RRN. Run robot-md register ROBOT.md before emitting any packets. See the Registry page for the full registration flow including key generation and RMN assignment.
For RCAN compatibility requirements, see rcan.dev/compatibility.