What evidence should teams keep for DSA recommender transparency?
Keep evidence showing that the public disclosure, live user interface, and recommender implementation describe the same system. For VLOPs and VLOSEs, DSA risk assessment, mitigation, audit, and data-access provisions can require explanations of algorithmic-system design, logic, functioning, and testing.
Use the product's actual criteria and records rather than an assumed scoring formula: the criteria that matter most, why they have their relative importance, the UI choices available to recipients, and the releases that changed the disclosure.
- Recommender inventory with surface name, owner, recipient group, ranking objective, main criteria, and whether the surface determines relative order or prominence.
- Terms-and-conditions extract showing the Article 27 main-parameter explanation and the options to modify or influence those parameters.
- UI screenshots, design specs, or QA evidence showing where each choice is directly available to recipients.
- For VLOPs and VLOSEs, evidence for the Article 38 non-profiling option for each recommender system and testing records for algorithmic changes.
- Change log tying recommender releases, terms updates, and interface changes to legal, product, and data-science review.
What should a DSA recommender transparency review verify before launch?
Before launch, verify that the platform has identified the recommender system, described the most significant criteria and their relative importance in plain language, disclosed recipient options in the terms and conditions, exposed any required choice control in the relevant interface, and, for VLOPs or VLOSEs, provided a non-profiling option for each recommender system.
Articles 34, 35, 38, 40, and 44 connect VLOP/VLOSE recommender systems to risk assessment, mitigation, data access, and choice-interface scrutiny.
Commission transparency page describes transparency reports, risk assessment and audit reports, and official reporting channels for VLOP/VLOSE practices including recommender systems.