Amidst a slew of stories of companies pivoting to ‘AI first’ models and reportedly selling user data to train Large Language Models the anti-AI crowd has received a rare win. After heavy backlash following a change to their terms of service, file sharing company WeTransfer has confirmed that user uploaded files will not be used as data to train generative AI models. 

According to BBC News, a recent change in the company’s terms of service included language that was interpreted to mean that files may be sold off to AI companies to improve the accuracy of their generative models, a move Facebook, Google and Twitter have all engaged in at various times. The specific wording in the contract included a clause stating that WeTransfer aimed to “improve performance of machine learning models that enhance our content moderation process” as well as a new protection which granted the company the ability to “reproduce, distribute, [and] modify,” files uploaded to the company. 

Naturally, this led to immediate backlash with many users claiming that they would boycott the company if they continued forward on this path. This comes after Dropbox, a company which offers a similar service to WeTransfer, dealt with a similar issue much earlier in the AI debate. 

In a statement, the company acknowledged that the change was originally made in order to open the door for the potential of using their service to train AI models but that they would no longer be pursuing this avenue. The new language in their ToS now states, “You hereby grant us a royalty-free license to use your Content for the purposes of operating, developing, and improving the Service, all in accordance with our Privacy & Cookie Policy.” Which still means that users who upload files to WeTransfer give up some of their rights to their files over to the company.