Transparency becomes an operational requirement
The European Commission has published guidance on the transparency obligations for providers and deployers of AI systems under Article 50 of the EU AI Act. The timing is immediate: the Commission states that Article 50 applies from 2 August 2026. For Belgian SMEs and any growing company serving people in the EU, this is a workflow deadline, not merely a legal reading exercise.
The guidance is intended to clarify scope, roles, definitions, exceptions, and ways to demonstrate compliance. It complements the Code of Practice on Transparency of AI-generated content. Businesses should confirm the legal interpretation for their own systems, but they can already implement the operational basics: inventory, ownership, notices, content marking, evidence, and change control.
What happened
According to the Commission, providers of certain AI systems must design them so people are explicitly informed when they interact directly with AI. Providers must also add machine-readable marking that supports detection of AI-generated or manipulated content.
Deployers have disclosure duties in specified contexts. The Commission highlights exposure to emotion-recognition and biometric-categorisation tools, deepfakes, and AI-generated text on matters of public interest when there is no human review or editorial control. The guidance also discusses concepts such as directly interactive systems, synthetic content, standard editing, and where exceptions may apply.
The purpose is practical: people should be able to calibrate trust and avoid deception, impersonation, fraud, and manipulation. The Commission notes that providers or deployers that do not follow the relevant voluntary code for marking and labelling will need to demonstrate compliance through alternative, equivalently adequate means.
Why it matters for growing businesses
Many SMEs are deployers rather than model providers. They buy customer-service bots, marketing tools, HR applications, meeting assistants, design software, and embedded AI features. That does not make transparency someone else’s job. The vendor may supply technical capabilities, but the business controls the context, user experience, content approval, and way outputs reach customers or employees.
A generic footer saying “we use AI” is unlikely to solve every use case. A customer needs a clear notice at the point of interaction. A synthetic campaign asset may need appropriate marking and disclosure. Public-interest content requires an editorial workflow. A material vendor or model update may change what the system produces and trigger reassessment.
Transparency also improves operations. Clear labels stop staff from mistaking drafts for approved records. Content provenance makes corrections easier. A documented owner can respond when a customer challenges an output. Machine-readable marks and retained source files strengthen traceability across agencies, channels, and languages.
Build one reusable disclosure workflow
Instead of asking every employee to remember the rules, add structured fields to the publishing or service process: AI involved, content type, human reviewer, required label, machine-readable mark, and approval date. Templates can then display the correct notice consistently on websites, emails, videos, or chatbot windows. The record should link to the final published asset and preserve any earlier generated version. This approach reduces omissions and gives management a clear view of which AI-assisted materials are live, who approved them, and when a legal or vendor change requires a new review.
Practical action checklist
- Inventory customer-facing bots, generated images, audio, video, public text, emotion analysis, and biometric categorisation.
- Record whether the business is acting as provider, deployer, distributor, or a combination for each use.
- Map where a person first encounters the AI system or generated content and place disclosure there.
- Ask vendors what machine-readable marking is supported and how marks survive editing, export, and publication.
- Define human editorial review for external content, especially material addressing matters of public interest.
- Preserve the original asset, generation record, approval, publication version, and later corrections.
- Test notices for clarity in every language and channel used by customers and employees.
- Add AI-transparency requirements to procurement, release checklists, and vendor-change reviews.
How VDS Consultoria can help
VDS Consultoria helps SMEs convert governance requirements into working systems. We can create an AI register, map content and customer journeys, configure approvals and evidence in Odoo, add disclosure steps to publishing and service workflows, and coordinate technical implementation with privacy, legal, HR, and communications teams.
This article is operational guidance, not legal advice. Qualified counsel should confirm the organization’s role, scope, exceptions, and final wording. The implementation challenge remains concrete: every required notice, mark, approval, and record must appear reliably in daily work before the deadline.
Source
European Commission, “Guidelines on Transparency of AI-Generated Content,” updated July 27, 2026: https://digital-strategy.ec.europa.eu/en/policies/guidelines-transparency-ai-generated-content