Legal experts say Meta’s AI may still scrape news sites
meta
| Source: The Verge | Original article
Meta revised its terms to drop language that implied it could train AI on news sites, but legal experts say the amendment fails to resolve their concerns.
Meta has revised its business terms to strip language that previously implied the company could train its AI models on content from news sites, including The Verge. The amendment removes the clause that suggested Meta’s systems were “associated with your products or services” when ingesting third‑party material.
Legal analysts, however, argue that the change does little to allay the concerns of publishers. One expert noted that the new wording “does nothing to restrict the ‘associated with your products or services’” language, meaning the underlying permission to scrape and repurpose news content remains effectively unchanged. The critique centers on the fact that, while the explicit reference to unrestricted training is gone, the broader terms still allow Meta to use publicly available data for model development, a practice that many news organisations view as a breach of copyright and a threat to their revenue streams.
The issue matters because AI developers increasingly rely on large, diverse datasets to improve large‑language models, and news publishers are pushing back against what they see as unlicensed exploitation of their work. If left unchecked, the practice could set a precedent for other platforms to sidestep licensing agreements, potentially reshaping the balance between content creators and AI firms.
Watch for possible legal actions from media groups, as well as responses from regulators in the EU and other jurisdictions that are tightening rules around data use for AI. Further revisions to Meta’s terms may be demanded, and the company could face pressure to negotiate explicit licensing deals with news organisations. The next few weeks are likely to reveal whether the amendment satisfies legal scrutiny or triggers a broader showdown over AI training data rights.
Sources
Back to AIPULSEN