The stalled New York bill demanding AI developers disclose every single URL in their model training roster might be a classic case of overpromising transparency but underestimating complexity. Sure, we all want responsible AI use and fair attribution, especially in journalism, but the proposal edged dangerously close to throwing a wrench into the gears of innovation.
At the core, large language models don’t work like a librarian filing away individual books with neat labels; they digest vast oceans of data and abstract patterns into math matrices—weights—making it nearly impossible to say exactly which URL led to which piece of generated answer. Expecting firms to maintain exhaustive logs of billions of URLs accessed and their precise influence is not just impractical but arguably irrelevant.
This reminds us how regulatory efforts must reflect the technology’s nature, not just good intentions wrapped in legal language. Mandating transparency without grasping data aggregation mechanics risks creating barriers that slow down progress while benefiting no one—not even the journalists seeking justice.
What’s more, recent legal rulings show courts gradually carving out fair use boundaries that recognize legitimate AI training activities, pointing toward balanced frameworks evolving organically. Instead of demanding raw data breadcrumbs, a smarter approach would be fostering dialogue between creators and AI builders to build trust and new licensing models.
So, yes, transparency matters, but let’s keep it real: demanding a digital paper trail crossing millions of clicks isn’t just impractical—it’s a bit like asking chefs to provide every ingredient origin on a recipe card for a complex stew. Bottom line? We need laws that encourage innovation and fairness without becoming a bureaucratic beast slowing the AI revolution. After all, tech progress works best when regulators and innovators speak the same language—and right now, that language is more math and less paperwork. Source: New York’s stalled AI bill would have blurred the line between disclosure and restriction
