Based on an article from iNFOnews, Canadian creators and publishers are urging the government to enforce transparency regarding the use of their content in training generative AI systems. These systems, which can produce text, images, and code, rely on vast amounts of existing content. Access Copyright claims that large language models profit from unauthorized use of copyrighted works in a "black box" manner, with creators unable to identify or control the use of their work.

Music Canada highlighted an incident where a fake AI-generated song mimicked the voices of Drake and The Weeknd, demonstrating unauthorized use of proprietary data. The Writers’ Guild of Canada called for basic disclosure obligations, emphasizing the knowledge imbalance between developers and creators. Some organizations have managed to sign licensing deals, but the Canadian Authors Association mentioned the immense obstacles rightsholders face due to lack of transparency.

Several lawsuits in the U.S. over similar issues underscore the need for clarity. The Canadian Media Producers Association stressed the urgent need for transparency requirements, as rightsholders struggle with evidence due to the destruction of training datasets. However, AI companies argue that such transparency is impractical. Microsoft and Google claim that recording and disclosing training data would inhibit AI development and expose sensitive information. Google added that existing copyright law exempts AI training, and a formal exemption would further support this.

Canadian AI company Cohere compared AI training to human learning, asserting it does not violate copyright. The Council of Canadian Innovators warned that disclosure requirements could disproportionately harm smaller companies and hinder the growth of the Canadian tech sector.