Recently, the social media platformXUpdated privacy policy will allow X platforms starting Nov. 15user datatrainAI Models, unless the user opts out, sparking user dissatisfaction. Previously, companies such as Adobe and Google have introduced similar content in their terms and conditions, causingAI TrainingControversies with privacy, copyright and other conflicts, and related legal issues are still under discussion.
Change: user data will be used for AI training
Recently, Platform X updated its privacy policy with a new clause that allows it to share user data with third parties to train AI unless the user opts out. However, the platform does not provide a clear opt-out option and reminds users that even if content is deleted on X, it may continue to exist elsewhere.
As previously reported, Platform X was sued by the Irish Data Protection Commission for violating the EU's General Data Protection Regulation by using its tweets to train the AI model Grok without notifying users. The lawsuit ended with X agreeing to stop using EU and EEA user data to train Grok.
And the updated privacy policy expands the scope of user data sharing to include third-party partners, sparking dissatisfaction among users, especially creators who worry about their works being used without compensation. Some users said that they have registered new accounts on other platforms and will transfer their works one after another. In response, Platform X responded that it would only use publicly available information to train AI models, and would not involve any private content. However, the definition of "public" and "private" content is unclear, raising questions.
Pitfalls: the legality of AI data training
Using user data to train AI models, X Platform is not an isolated case. Over the past year, Adobe, Google, Zoom and many others have introduced the content into their terms of service or privacy policies.
Social platforms provide a rich source of data for AI models, including conversational content, slang, and real-time news, all of which are important elements in AI training. But Susan Schreiner, technology and social media analyst at C4 Trends, also noted that "the quality of the training data is as important as the data itself in a generative AI framework. If Platform X's AI Grok is trained on hate speech or worse, the AI is prone to replicate and generate misleading content."
Meanwhile, the use of user data to train AI involves copyright and privacy issues, and in the existing relevant regulations, whether the user has given informed consent is one of the criteria for judging the legality of the act. According to the EU's General Data Protection Regulation, companies must obtain explicit consent from users to use their personal data for AI training. California's California Consumer Privacy Act (CCPA) states that users have the right to know how companies use their data and have the option to refuse their data being sold or used for AI training.
Many large companies use terms of service or privacy policies to seek users' consent in order to avoid legal risks, but there are still some problems. For example, the language of the terms and conditions is too complex or vague, so it is difficult for ordinary users to understand the specific meaning; many of the terms and conditions need to be "accepted in full", and the user can only agree to use the platform service, which is in a passive position; the lack of a clear withdrawal mechanism to revoke the previous "consent". Regarding the legality of the data used for AI training, the relevant laws and regulations still need to be further improved.