1. Where AVC uses AI

AVC may use AI models behind agents, assistants, research tools, content generation, classification, planning and decision-support workflows. Some deterministic platform functions do not use AI. The presence of automation alone does not mean every AVC operation is AI-generated.

2. Direct interaction with AI

When an AVC AI system directly interacts with a natural person, the product should make the AI nature of that interaction clear from the start where required, unless it is already obvious in the circumstances. AI identity must not be hidden behind wording that makes an autonomous system appear to be a human operator.

3. Current EU transparency framework

The EU AI Act Article 50 transparency obligations apply from 2 August 2026 to the systems and deployments that fall within their scope. AVC therefore treats AI-interaction disclosure and applicable generated-content transparency as production requirements, while still assessing each concrete use case rather than assuming one generic badge satisfies every obligation.

4. Governed agents

AVC agents operate through scoped capabilities, policy, risk boundaries and explicit authority. An AI model's confidence does not itself create permission to send a message, spend money, change protected state or perform another consequential action.

5. AI-generated and manipulated content

AVC must apply the relevant transparency and machine-readable marking requirements when its role, content type and deployment fall within those obligations. Deepfakes, public-interest text and other regulated categories require separate treatment rather than one generic "AI generated" badge being assumed sufficient for every case.

6. Limitations of AI output

AI output can be incomplete, outdated or incorrect. AVC therefore distinguishes generated recommendations from verified evidence and should expose meaningful provenance, approval and maturity information for consequential workflows.

7. Human ownership and control

Jonas Abde remains the legal owner and ultimate natural-person authority for AVC. Agent autonomy is bounded delegation inside the software system, not independent legal personality or a transfer of legal responsibility to an AI model.

8. Models and providers

AVC can use multiple model providers and may route work according to capability, policy, privacy, cost and reliability requirements. Provider selection may change over time. Product-specific notices should identify material provider or processing information where required for that use.

Compliance boundary

9. No blanket compliance claim

This notice is a transparency surface, not a certification that every AVC model, agent, output or future Product Cell satisfies every obligation of the EU AI Act. Each deployed AI use case must be classified and verified against the obligations that actually apply before production approval.