New EU laws make AI content labels compulsory – but might just make it harder to spot deepfakes
Labels alone won’t be enough to tackle the rise of deepfakes.
The European Union has recently enforced new transparency obligations for tech companies, requiring them to clearly label AI-generated content that appears authentic. This marks a significant step up from previous recommendations made by other countries, such as Australia, which only advise AI developers to label AI-generated content.
Research suggests that our ability to distinguish between fake and real content when confronted with sophisticated AI-generated material is often no better than chance. However, the effectiveness of the EU's approach remains uncertain and could potentially introduce unintended consequences. The EU AI Act, which took effect on August 2, introduces various transparency obligations for companies that provide or deploy AI systems or content.
This includes disclosing AI-generated or manipulated images, audio, or videos that resemble existing persons, objects, places, entities or events, as well as AI-generated or manipulated text on matters of public interest that is not subject to human review or editorial control. Iconic labels are to be used to indicate AI-generated content, with specific requirements for audio deepfakes and content that is evidently artistic, creative, satirical, or fictional.
Companies are also required to embed invisible, machine-readable watermarks on any content generated or processed with their systems. Compliance with the act is mandatory for companies, though non-compliant entities may face substantial fines. However, the act does not apply to individuals when they create content in a personal or non-professional capacity.
While the EU's labels may help people identify synthetic content, their effectiveness could be undermined by unintended consequences. For instance, over-reliance on labels might lead people to assume that content without a label is genuine, making them more susceptible to undisclosed and malicious AI use. Moreover, labels are just one of many cues people use to determine the trustworthiness of online content, and their effectiveness depends on factors such as the timing and presentation of the labels, as well as the social cues surrounding the content.
Cultivating AI literacy and critical evaluation skills among users is essential to counteract the potential limitations of AI-generated content labels.
Written by urgent.news from The Conversation AU's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
