What a company must do to meet AI Act transparency obligations by 2 August 2026

Immediate actions: decisions and deadlines

Companies must implement mechanisms by 2 August 2026 to inform users when they interact with an AI system or when content was generated/altered by AI, per European Commission guidelines (digital-strategy.ec.europa.eu). The first actions are to assign internal responsibility (product owner, compliance, IT) and include the compliance deadline in the project roadmap.

How to identify covered systems and classify them

Organizations should inventory AI systems and classify each as: interactive AI (direct user interaction), content generator (text/image/audio), or assistive tool. Classification determines required transparency measures:

1. For interactive systems: visible UI notice or verbal cue that the user is interacting with AI.
2. For generated content: a label/mark next to the content and metadata indicating origin.

Classification and the inventory must be recorded to support audits and reporting.

How to prepare systems technically and procedurally

Technical and procedural measures must proceed in parallel:

1. Software changes: embed UI banners/notices, audio cues, add provenance metadata and content markers; consider watermarking or industry provenance standards (e.g., C2PA) as technical options.
2. Logging and evidence: implement logs that demonstrate the disclosure was shown to the user and record content provenance.
3. Organizational processes: update privacy policies, train product and support teams, and add transparency clauses to vendor/model contracts.

Every technical choice should be documented in the compliance package and linked to the system risk assessment (see AI Act text available on EUR-Lex).

Where the main risks and pitfalls lie

Key risks include inconsistent notices across channels, tamperable or removable labels, and reliance on third-party models without provider guarantees. Labels must be understandable to average users-overly technical wording risks allegations of deception. Legally, the obligations apply within the EU jurisdiction and stem from the AI Act and the Commission guidelines; companies operating in the EU must reflect these requirements in contracts and processes.

Summary

By 2 August 2026 companies should have an AI inventory, implemented user notices and content markings, logging mechanisms, and updated processes and contracts. Technical details should be agreed with legal and operational teams, following the European Commission guidance and the AI Act text on EUR-Lex.

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.