Intellectual property implications AccuracyINSIGHT 4 INSIGHT 5Accuracy and bias: GenAI outputs can amplify existing biases in training data and produce discriminatory outputs or generate hallucinations (false information), which can erode consumer trust and content value. Poor representation of diverse communities in model training datasets increases algorithmic bias, risking the marginalization of certain voices and reduction of models’ accuracy.49 Addressing these challenges is essential for industry leaders and society to ensure AI’s sustainable and responsible adoption. It calls for a human-centric and holistic approach. As regulation varies and evolves at a different pace across regions, there is opportunity for industry self-governance to complement regulation (“co-regulation”) through methods like collective bargaining, binding commitments, best practices and voluntary standards. Key governance themes have been defined and examined as part of the World Economic Forum’s Digital Trust Framework.51 They include: –Accountability: Organizations are defining principles for responsible AI adoption. Some are establishing governance bodies, like supervisory boards and AI councils, as well as human oversight processes to ensure ethics and transparency standards are upheld. These committees should consider diverse perspectives from technologists, ethicists, legal experts, creators and others to effectively assess genAI products and features. They should be responsible for reviewing AI practices, identifying potential risks and ensuring compliance with both internal policies and external regulations. Defining evaluation frameworks based on different categories of AI models, data sources and use cases, along with cross-industry standards like ISO-4200152 can streamline review and approval processes. –Fairness: Companies should use AI models that minimize bias and mitigate unintended consequences in content creation and distribution. This will ensure equitable treatment, inclusivity and fairness across content platforms while safeguarding user data rights. –Transparency: Nurturing consumers’ trust requires organizations to inform about AI- generated content and its use through appropriate labelling and disclosures within the product experience – whether auto-generated, auto-generated with human oversight or human-generated. Information on related data practices, safety policies and potential risks (such as bias and privacy) of the AI model used in genAI products should be made available via accessible documentation. Standards and technical solutions to ensure content authenticity, such as digital watermarking, content origin and history, and blockchain- based rights management, are currently under development to support a trustworthy information ecosystem. However, successful adoption at scale requires policy frameworks that are aligned with common principles, rules and technological standards.53 Companies should use AI models that minimize bias and mitigate unintended consequences in content creation and distribution. A recent survey found that 52% of respondents view IP infringement as a significant risk. Research highlights a growing concern around genAI inaccuracy, with 63% of respondents considering it a relevant risk.25% report actively working to implement measures to mitigate it.48 38% declared that they are working to mitigate it.50 Artificial Intelligence in Media, Entertainment and Sport 19

Artificial Intelligence in Media Entertainment and Sport 2025