Sony Music warns AI firms over ‘piracy’ of content
Sony Music has written to more than 700 artificial intelligence companies, accusing them of using its artists' content without permission, as tension builds between the creative and technology sectors.
The world's second-biggest record company has asked technology groups including Google, Microsoft and Meta to reply within weeks, providing details of lyrics and songs used to train their technology, and how they had been obtained.
Sony said this was with a view to discussing licensing or partnership deals to pay for content, but it is understood that it could pursue legal action if companies do not comply.
"Due to the nature of your operations and published information about your AI systems, we have reason to believe that you and/or your affiliates may already have made unauthorised uses of Sony Music Group content in relation to the training, development or commercialisation of AI systems," the letter said.
Sony said it demanded the disclosure under the terms of the forthcoming European Union AI Act, which requires companies making AI models to reveal the content used for training "and to put in place a policy to respect EU copyright law".
The business is understood to already be in discussions with 350 technology companies about partnerships and licences.
Sony also warned companies about any future use of its copyright.
It said it would be "open to discuss licence agreements" but that its "well-established position" was that permission was required in advance.
Everything from lyrics to images to compositions was protected, the label said, and unauthorised use deprived artists "of control over and appropriate compensation" for their work, and "infringes our intellectual property and other rights".
Sony will expand its list of AI companies that are in the scope of its ban, and has told online platforms and services that distribute and stream music digitally, such as Spotify, to make it "abundantly clear" to users that mining and training on its content is not permitted.
Generative AI models require vast amounts of data in order to train their technology in pattern recognition, which enables the creation of music, images and text mimicking human-created material, as made by the likes of ChatGPT and StabilityAI.
However, there is increasing concern in the creative industries, from musicians to publishers, that their data may have been used without permission to train AI.
Copyright use is proving a headache for policymakers, who are trying to tread a line between supporting innovation and protecting the creative sector.
Britain has failed to agree a satisfactory plan after talks between the two sides, convened by the Intellectual Property Office, collapsed. The task of creating guidelines has been picked up by the Department for Science, Technology and Innovation, but nothing has yet been published.