Article 50 — transparency
Art 50(1) — chatbot disclosure: GenZAgents doesn't implement the runtime disclosure (that lives in your UI), but we capture the disclosure event as a receipt field. Art 50(2) — AI-generated content labelling: receipt ID becomes the watermark anchor (C2PA adapter on Enterprise tier). Art 50(3) — text content labelling: per-receipt reviewer_human_id distinguishes human-reviewed from purely-AI. Art 50(4) — verifiable authenticity: signed receipts ARE the verifiable format.
Article 12 — high-risk system record-keeping
High-risk AI systems must maintain "logs that allow tracing operations and verifying compliance". The receipt feed is exactly that. The Article specifies the logs must allow assessment of: how the system operated, which data it processed, what decisions it made, who the operators were. All five are captured per receipt.
Article 9 — risk management
You operate the risk management system; receipts evidence its operational execution. When the auditor asks "show me the risk mitigation in action", filter receipts to show the relevant operational decisions + supervising humans + outcomes.
Article 10 — data governance
You document the data governance; receipts evidence per-call data handling. Filter receipts by the data class involved; show provenance + access controls + retention.
When the AI Act bites
Article 50 applies from 2 August 2026. High-risk obligations (Articles 8-15) apply from 2 August 2027 (for most use cases). The pragmatic deadline for an operational receipt-based evidence layer: mid-2026 — which is exactly the GenZAgents launch window. Early adopters get the operational evidence baseline that competitors will struggle to retrofit.
Penalty exposure
AI Act non-compliance: up to €15M or 3% of worldwide turnover for Article 50 violations. For a £100M turnover company: £3M of risk per audit. GenZAgents Enterprise at £6k/year is rational risk management at any scale where Article 50 applies.