欧盟AI版权新规引发创作者与技术巨头激烈博弈
When the European Union’s Artificial Intelligence Act finally limped past its final legislative hurdles in early 2024, the applause from Brussels was somewhat muted by a single, contentious measure buried deep within its text. Article 53, a late addition, compels developers of general-purpose AI models to publish a ‘sufficiently detailed’ summary of the copyrighted material used in training. It represents the world’s first binding transparency requirement for training data, and it has cleaved the continent’s creative and tech sectors into warring camps. Far from a mere technicality, this clause has reignited a philosophical battle over whether innovation should be allowed to consume the fruits of human imagination without consent, redefining the meaning of intellectual property in an algorithmic age.
The provision emerged from a protracted tug-of-war between the European Parliament, which sought a strict opt-out system for rightsholders, and the Council, which hewed closer to the industry’s desire for minimal friction. The final compromise stops short of requiring individual licensing, but it demands that AI firms disclose their sources — a half-measure that has pleased almost no one. Publishers, music labels, and visual artists contend that mere transparency is little more than a gesture; without a mechanism to enforce their opt-outs, they argue, the Act becomes a ‘pirates’ charter’. Meanwhile, technology companies warn that the obligation, however vague, could hobble European start-ups by forcing them to navigate a legal minefield that their American and Chinese competitors simply bypass.
Vocabsavvy AI · a level-headed international affairs editor · Vocabsavvy Original