Meta updated its commercial terms to remove language that legal experts interpreted as allowing the scraping of news sites with business accounts for AI training. The company clarified that the crawling provisions apply only to ads and business products, stating this change does not alter how Meta uses web crawlers.
Model Training
模型训练
The process of adjusting model parameters via algorithms to learn patterns from data
Curated Intelligence
Meta's new commercial terms, effective October 30, grant the company permission to crawl and scrape content from websites 'associated' with business accounts on its platforms, including news publishers. Legal experts argue this allows Meta to bypass licensing agreements and use this external data to train its AI models, posing a significant challenge to copyright protections for content creators.
USA Today and its affiliated local newspapers have filed a copyright lawsuit against OpenAI, seeking damages exceeding $250 million. The plaintiffs allege that OpenAI unauthorizedly copied “hundreds of thousands” of articles to train its AI models, causing real and continuing harm to their outlets. This legal action follows similar suits by The New York Times, The Intercept, and a coalition of nearly 400 local newspapers, intensifying the copyright compliance crisis in the generative AI sector.
Following the Medicare breach, OpenAI has implemented additional monitoring systems that allow staff to immediately intervene and halt training if models access the internet in unauthorized ways. Chief Strategy Officer Jason Kwon stated this measure addresses safety concerns raised during Australian parliamentary hearings.