File-sharing platform WeTransfer has reversed a change to its terms of service after a clause appeared to give the company a perpetual, worldwide licence to use customer files to “improve machine-learning models”. The wording, introduced this week in section 6.3 of the agreement and due to take effect on 8 August, triggered widespread criticism from creative professionals and privacy advocates who rely on the service to transfer large, often copyrighted, material. Within two days of the backlash, WeTransfer issued a statement to outlets including the BBC saying it has never used user content to train artificial-intelligence systems and does not sell data to third parties. The company republished its terms, stripping all references to AI and machine learning and limiting the licence to operating, developing and improving the service in line with its privacy policy. The incident follows WeTransfer’s 2024 sale to Italian tech investor Bending Spoons, which has since sought to boost the platform’s profitability. Co-founder Ronald Hans, no longer involved with the business, said the episode “betrays” users and announced plans for a competing file-transfer service called Boomerang that he says will prioritise creator rights. WeTransfer, used by about 80 million people a month in 190 countries, is the latest technology firm to draw scrutiny over how customer data might be employed in AI development. The swift climb-down underscores rising regulatory and reputational risks for companies perceived to overreach in harvesting user content.