Meta Removes Terms Language That May Have Permitted News Site Scraping for AI Training

Key Change to Platform Terms

Meta has removed language from its terms of service that experts say may have enabled the scraping of news sites to train its artificial intelligence models, according to reporting by Terrence O'Brien.

What the Language Allowed

The removed clause, according to technology and legal experts, could have been interpreted as granting Meta broad permission to collect and use content from third-party news publishers for AI training purposes. The change quietly alters how the company's user and platform agreements address automated data collection.

Why It Matters in 2026

The revision arrives amid heightened scrutiny of how major AI developers source training data. Throughout 2025 and into 2026, publishers, regulators, and courts have increasingly challenged unlicensed scraping of news content, and several jurisdictions have moved toward stricter disclosure and licensing requirements for AI training datasets. Meta's adjustment reflects growing pressure to clarify—or narrow—the permissions embedded in platform terms.

Implications for Publishers and AI Development

The removal does not, on its own, resolve ongoing debates over fair use, licensing, or compensation for news content used in AI systems. However, it signals that platform terms are becoming a focal point in the broader fight over data provenance and publisher rights as AI models continue to scale.

via The Verge AI