The AI platform in the EU faces a compliance test
The EU has officially classified ChatGPT, Reddit and Roblox as super-large online services and has applied the strict rules of the DSA.
AI chatbots are also included among the rules listed by the DSA.
The European Commission has published the latest results of its determination. All three services have surpassed the threshold of 45 million monthly active users in the EU region. The Reddit forum and the game community Roblox were designated as super-large online platforms, while the online search engine ChatGPT was designated as a super-large online search engine. Once this scale is reached, the platforms must assume the highest level of responsibility under the DSA framework.
Many people think that DSA is only for social websites or e-commerce platforms. This determination represents that the EU has officially placed generative AI chatbots into this mature digital regulatory system. Previously, X's Grok has already been undergoing relevant DSA investigations. Now ChatGPT has also joined in, and AI chat tools cannot be simply regarded as a software product. They need to take responsibility for user-generated content and information dissemination like traditional large platforms.
After receiving the notification, the three companies have a time window to complete the rectification. All new compliance obligations need to be implemented by the end of December of this year. OpenAI has already started preparations for its outward appearance and team to complete the new compliance work. Roblox replied relatively positively saying it is the first gaming platform in the EU to reach this milestone and safety is the bedrock of product development. It is willing to continue communicating with EU institutions. Reddit has not responded to media inquiries for the time being.
The platform is responsible for the content and minors, and the AI bill and DSA form a dual constraint.
After becoming a super large online platform or search engine, the most essential requirement is to promptly identify and delete illegal content within the platform. We also need to establish specialized mechanisms to protect the privacy and online safety of minors. When applied to ChatGPT, this requirement is different from that of regular social platforms. Most of the content on Reddit and Roblox comes from user posts and uploaded works. The output content of ChatGPT is generated by an AI model. According to the DSA rules, the platform also needs to evaluate the risks brought by AI output content and handle the violation information generated by the model.
Many people easily confuse the EU DSA and AI bills. The focus of these two sets of laws is different and will simultaneously affect these platforms. DSA is more inclined towards platform operation and content distribution. Regardless of whether the content is posted by humans or generated by AI, as long as it is disseminated externally through this service, the platform must bear the responsibility of content management. The AI bill focuses on the AI model itself, setting rules for model training, capabilities, and risk grading. On the one hand, the EU implements DSA to control the flow of information on the platform. On the other hand, implementing the world's first AI specific regulations to standardize AI technology research and development.
Prior to the implementation of this regulatory measure, the United States had repeatedly raised concerns that the EU's digital regulations specifically target American technology companies and restrict freedom of speech. Especially after X platform was fined by the EU, the related debate has become more intense. But the position of the European Commission is very clear. The rules of DSA apply to all companies providing services in the European Union, registered in the United States, China, or other countries, with completely consistent standards.
For AI products, compliance costs will reshape the product design logic.
For AI enterprises like OpenAI, the changes brought by DSA are not just the addition of an auditing team. The underlying logic of the product also needs to be adjusted accordingly. Generative AI like ChatGPT inherently has the problem of uncontrollable content. The same prompt can result in different output results each time for the model. To meet the requirements of DSA, OpenAI needs to upgrade the content filtering and risk warning mechanisms. Such modifications will increase technical costs. The model interaction experience may also be affected. The challenges faced by Reddit and Roblox mainly come from the massive user content. Reddit has numerous community sections, and the content in different sub-boards varies greatly. On Roblox, there is a youth group, and the number of user-created game content is huge.