Disagreement over AI transparency code to focus on 'deceitful intent'

Proposed rules for labelling AI deepfakes under the EU Artificial Intelligence Act (AI Act) have sparked disagreement between industry stakeholders and civil society groups. The legislation requires companies to watermark AI-generated content, such as images, audio, and video, to prevent public deception. However, defining the threshold for what constitutes a necessary label has proven contentious during the drafting of a transparency code of practice.

The initial draft of the code, published before Christmas, favours a broad interpretation supported by civil society. This approach would require labels for even minor AI enhancements, such as noise removal in audio, if the edits alter the content's context. Industry representatives argue that such all-encompassing requirements would lead to over-labelling, potentially diluting the impact of warnings and placing an undue burden on sectors like advertising. A central point of debate is whether labels should only be mandatory when there is demonstrable deceitful intent.

Disputes also extend to the watermarking of AI-generated text, including software code, which industry sources claim could reduce technical quality. Although the transparency code remains voluntary, adherence is expected to serve as a benchmark for formal compliance. While the transparency rules were initially scheduled for August 2026, the European Commission has proposed a delay until February 2027.

Training Announcement: The BCS Foundation Certificate in AI examines the challenges and risks associated with AI projects, such as those related to privacy, transparency and potential biases in algorithms that could lead to unintended consequences. Explore the role of data, effective risk management strategies, compliance requirements, and ongoing governance of the AI lifecycle and become a certified AI Governance professional. Find out more.