Pilot
€0 / 30 days
For 1 deployment, 1 monthly evidence run. Validates the loop before commitment.
Start pilot →Compliance · 2026 ready
Monthly evidence packs for Art. 26(4)-(7) + Art. 50. Pure logic, no LLM in the evidence path. Audit-ready by design.
What the loop covers
EU AI Act Article 26 imposes five deployer obligations for high-risk AI systems. Our evidence loop evaluates each one against your operational evidence sources and emits a byte-identical, replayable evidence pack. Pure logic. No LLM in the evidence path — which means your auditor can replay the exact same computation and get the exact same result.
Art. 26(4) — Transparency
Deployer shall inform natural persons that they are interacting with an AI system, unless obvious from context.
Evidence: UI-component register, AI-disclosure policy version, runtime notification log per deployment.
Art. 26(5) — Human Oversight
Deployer shall ensure human oversight by trained, authorized personnel during AI system use.
Evidence: Oversight-role assignment log, intervention timestamps, override-rate per deployment.
Art. 26(6) — Incident Monitoring
Deployer shall monitor AI system operation for risks and report serious incidents to providers and authorities.
Evidence: Incident classification, response timestamps, regulator-report trail.
Art. 26(7) — Input Data Quality
Deployer shall ensure input data is relevant, sufficiently representative, and free of errors.
Evidence: Input-data inventory, data-quality metrics, drift-detection log.
Art. 50 — User Notification
Providers shall inform users that AI-generated content is AI-generated, including emotion-recognition and biometric-categorization systems.
Evidence: Disclosure-banner configuration, AI-content marking protocol, provider-to-user notification log.
All five — in one evidence pack
One monthly run. Byte-identical output. Replayable on demand. Your auditor pulls the evidence, replays the loop, gets the same evidence pack. No "trust me, the LLM said so."
For developers
Step 1 — Connect your evidence sources
Webhook to your existing logs (incident response, oversight interventions, input-data quality). One integration per obligation. Set-and-forget.
Step 2 — Schedule the monthly run
Cron-triggered. UTC-aligned month boundaries. One evidence.json per deployment per month, byte-identical, replayable.
Step 3 — Hand to your auditor
One signed evidence pack. Includes reproducibility hash + replay command. Auditor replays locally, gets the same result.
POST /v1/audit-bazaar/clone/audit-bazaar-eu-ai-act/mcp/run_eu_ai_act_deployer_evidence
{
"deployment_id": "dep_support-triage-prod",
"scope": "2026-07",
"sources": [
{"id": "incident_log_prod", "type": "incident_response_webhook"},
{"id": "oversight_log_prod", "type": "human_oversight_trail"},
{"id": "input_quality_metrics", "type": "data_quality_dashboard"}
]
}
Pricing
For 1 deployment, 1 monthly evidence run. Validates the loop before commitment.
Start pilot →For 1-5 deployments, monthly evidence runs, auditor handoff package, signed reproducibility hashes.
Choose Compliance →For 5-25 deployments, weekly evidence runs, S3 archive, audit-log integration, Slack support.
Choose Ops →25+ deployments, multi-region, custom obligations, ISO 42001 + SOC 2 evidence packages, dedicated compliance specialist.
Talk to us →Annual contracts save 20%. EU AI Act deadlines: high-risk system deployers under scope from 2 August 2026 — pilot now to be ready.
Trust ledger
Byte-identical
Same input → same evidence pack. Replay locally → same hash. No "trust me."
Pure logic
Zero LLM in the evidence path. Deterministic, deterministic, deterministic.
Simulator-validated
Alpha-grade passes the simulator regression. Production is the next phase — see the proof-points ledger.
5-min setup
Webhook your evidence sources, schedule the run, hand the package to your auditor. Half a day, not a quarter.
Multi-regulation
EU AI Act + ISO 42001 + SOC 2 evidence packages from the same underlying loop.
Every claim above is verified in docs/proof-points-ledger.md. Pending claims (PDF <5MB, byte-identical PDF, S3 integration, hours-saved, production-ready) are flagged until the row flips to verified.
Get started
EU AI Act high-risk deployer obligations are enforceable. Run a free 30-day pilot — get a real evidence pack for one deployment, one obligation set, end-to-end.