Due to lack of dataset (to our knowledge) to benchmark detection of machine-generated text, we aim to provide a multilingual dataset (French, Spanish, English) to benchmark text on information manipulation by generating news-like texts and short texts like microblogging.