LI Wenlong
Research Professor, Guanghua School of Law, Zhejiang University
The authentication of AI-generated content is fast shaping up to be one of the central legal challenges in global AI governance. Four main technical approaches have been put forward for AI-gen content authentication—namely post hoc pattern detection, retrieval-based detection, content provenance (metadata based), and digital watermarking—each of which comes with its own strengths and drawbacks. For now, regulators tend to fall back on a combination of the latter two methods. As synthetic content grows ever more lifelike, the line between human- and machine-generated material is steadily blurring, and the gap between legal frameworks and real-world developments is becoming harder to paper over. Jurisdictions such as China, the United States, the European Union, and Brazil have all rolled out relevant rules, though they differ in where they place the emphasis. At the global level, existing international bodies and forums have struggled to step up and drive the issue forward. Meanwhile, the rollout of technical standards such as C2PA continues to run up against both technical hurdles and political pushback (both domestically and internationally). What is urgently needed is to bring about cross-border, cross-platform cooperation on content labelling; only by pulling in the same direction can stakeholders hope to stamp out the societal risks posed by disinformation, deepfakes, hallucinations and the sheer volume of AI-generated content.
