From 2 August 2026: new AI transparency obligations for chatbots and generated content

Introduction: what changes on 2 August 2026

From 2 August 2026 providers and implementers of AI systems must comply with new transparency obligations set by the European Commission; the rules require disclosing when users interact with AI, labelling content generated or altered by AI, and applying machine-readable marks that make such content easier to detect, according to the European Commission.

Direct impact on AI providers and implementers

The obligations apply directly to chatbots and other interactive AI: systems must clearly inform users they are interacting with AI rather than a human. Deepfakes (images, video, audio edited or generated by AI) must be labelled, and AI-generated or altered content must carry machine-readable markings - these are the core requirements listed by the European Commission. For businesses this means changes to UX, metadata handling and the technology supply chain.

How to prepare systems and compliance processes

1. Update user interfaces so disclosures that an interaction involves AI are clear, persistent and present across all customer touchpoints.
2. Implement processes to attach machine-readable metadata to all AI-generated or AI-modified content; per the Commission, these marks must allow automated detection.
3. Establish audit and retention procedures for markings and logs, and record model versions and decision traces to demonstrate compliance during inspections.
4. Amend contracts with model vendors and third-party service providers to secure marking obligations and metadata access.

Compare implementation details against the AI Act text available on EUR-Lex and the Commission communication to determine the precise scope of obligations.

Where the risks and pitfalls lie during rollout

Main risks include inconsistent labelling across channels, failure to mark third-party modified content in the supply chain, and insufficient evidence trails for audits. Multilingual systems and integrations with external APIs may complicate machine-readable marking. Enforcement will be carried out by national authorities together with the Commission's AI Office, so delayed or incomplete action can lead to enforcement measures, according to the European Commission.

Conclusion

From 2 August 2026 organisations must rapidly implement visible AI disclosures, label deepfakes and add machine-readable metadata. Prioritise UX updates, metadata pipelines, auditability of model versions and vendor contract clauses. Review compliance processes against the Commission communication and the AI Act text on EUR-Lex without delay.

Lub System helps B2B companies implement AI, automation and IT solutions end-to-end - from strategy to deployment. See our services or get in touch to discuss your case.