Meta, Facebook and Instagram’s parent company has revealed through Nick Clegg, its President of Global Affairs, its strategy for training its new Meta AI virtual assistant.
According to Meta, it uses public Facebook and Instagram posts as a part of the training data for its Meta AI. However, it excludes private posts shared exclusively with family and friends, while claiming to respect consumers’ privacy Furthermore, private chats within Meta’s messaging services were not used for training, and the company went to great lengths to filter private details from public datasets employed in the training process.
Clegg stressed their careful selection of datasets, focusing on publicly available information while avoiding those containing a heavy preponderance of personal data. He cited the decision not to include LinkedIn data due to privacy concerns.
Clegg also highlighted safety measures, including restrictions on creating photo-realistic images of public figures. He anticipated potential litigation related to copyright issues and fair use doctrine, indicating that this legal aspect may unfold in the future.
The sources for this piece include an article in Reuters.