The Digital Services Act (DSA), along with the Digital Market Act (DMA) are initiatives from the European Union Commission, proposed in December 2020 and agreed upon in April 2022. The main goal is to provide and ensure an accountable online environment in the EU, and regulate the “gatekeepers” on online interactions.
This is a wide-reaching act that will differentiate “gatekeepers” from other online services, and apply obligations to them to create a more fair, competitive, and safe online market. These acts are not in full force yet – they are subject to formal approval by the European Parliament and European Council, and will apply six months after adoption.
Which Organizations Are Impacted?
The DSA makes a clear demarcation between small and medium-sized organizations and large online intermediaries and platforms: The large intermediaries and platforms will carry more responsibility and accountability to both allow freer competition and access to information, and to be transparent in their content moderation decisions. SMEs that provide digital services in the EU (regardless of where they are based) will also be subject to the DSA, but with “obligations proportionate to size, impact, and risk”, as stated by the Commissioner for the Internal Market Thierry Breton.
One of the most labor-intensive actions that online platforms, particularly those that reach 10% of consumers in the EU and subject to the strictest obligations, will need to take is to provide transparency into the algorithms used for content moderation choices, recommendations, and many other functions. However, this could have some of the largest impacts on several fronts:
- In beginning to engage with the societal influence of machine learning algorithms – not addressing harms done to individuals, but influences shaping the way large populations of people perceive the world
- In enforcing accountability on proprietary algorithms, the handling of which will set a precedent for transparency of other proprietary algorithms used in other areas of business
- In stimulating more investment in explainability for machine learning models, which impacts the transparency of models, the ability for human administrators and those impacted by model recommendations to understand those recommendations, and boosting security of models by making it more human-understandable if a recommendation has been maliciously influenced.
This could be a turning point in our ethics debate on machine learning, moving from the theoretical to the practical and enforceable. This will be closely followed by analysts at KuppingerCole Analysts, as well as the numerous other impacts including the way that businesses can connect and communicate with their consumers or future customers. We will continue to assess and prepare you for those ongoing transitions.