Attempts to agree on a UK voluntary code of practice for training artificial intelligence (AI) have been abandoned by a group of industry executives including from Microsoft, DeepMind, IBM, the BBC, British Library and Music Publishers Association.
The group was convened by the UK's Intellectual Property Office of AI software developers and music and media companies.
A code of practice is sought to govern how AI models are trained using copyrighted materials such as newspapers, books, music, photography, art and other products of human creativity, which has caused concern within the creative sector and already led to several lawsuits in the US, such as several authors and the New York Times.
The Intellectual Property Office's efforts to formulate a voluntary code of practice for text and data mining have stalled due to disagreements among a group of industry executives, according to a report from the Financial Times, which is one of the media companies in the IPO working group.
One of the sticking points for media and creative side was transparency on what has been used so far to train AI models, though this information has not been provided.
As the working group, which also includes representatives from the Alliance for IP, Creative UK, the English Premier League, Professional Publishers Association, RELX PLC, and UK Music, has been unable to reach a consensus, the responsibility has shifted back to the Department for Science Innovation and Technology, which was an observer on the talks.
The delay comes amid growing worries among artists and authors about their work being used without compensation.
A forthcoming white paper is expected to outline the UK's AI regulation strategy but may not provide concrete policies, the report said.
The BBC, NYT, CNN, and Reuters are among several publishers that have blocked AI developers like ChatGPT developer OpenAI from unauthorised crawling and scraping of their online content.
OpenAI has offered newspapers and other publishers between US$1 million and US$5 million per year for licences to gain access to their articles to train its large language artificial intelligence (AI) models.