Meta's Clause Revision Still Fails to Stop AI Scraping? Legal Experts Say News Sites' Risks Persist

Meta recently revised its business terms, removing language that implied it could freely use news website content to train AI models, but legal experts point out that this adjustment does not truly resolve news publishers' concerns. According to The Verge, the phrase "related to your product or service" remains in the terms, potentially continuing to leave room for Meta to scrape news content.

Clause Revision: Superficial Concession or Substantive Evasion?

Meta removed certain wording that could be interpreted as allowing it to use news website data to train AI without restriction, appearing to respond to publisher pressure. However, legal experts argue that the revised terms still retain a key loophole—the phrase "related to your product or service." This wording is extremely broad and could be interpreted by Meta as meaning that as long as scraping behavior has some connection to its own products or services, it falls within the scope of compliance.

This means that even if news websites explicitly prohibit AI scraping through robots.txt or terms of service, Meta may still invoke this clause as a defense. The clause revision appears more like a public relations gesture than a substantive legal constraint.

robots.txt is a text file located in a website's root directory that tells web crawlers which pages can be crawled and which cannot. It belongs to the "Robots Exclusion Protocol," an industry self-regulatory convention rather than a legal mandate. Websites configure robots.txt to prohibit AI crawlers from accessing their content, but compliance depends entirely on the voluntary cooperation of crawler operators. This means that even if news websites explicitly prohibit scraping, companies like Meta may continue scraping on the grounds of contractual authorization, and robots.txt cannot provide legal protection. This is one reason why legal experts consider the clause loophole more critical than technical measures.

Core Doubts from Legal Experts

Experts' concerns center on two points. First, the ambiguity of the clause gives Meta excessive interpretive authority. It is common practice for tech giants to bury broad wording in user agreements, and publishers lack equivalent bargaining power to remove such clauses when signing. Second, the current legal framework lacks clear definitions regarding the scraping of AI training data. The "fair use" principle in copyright law is highly contested in the AI context, and Meta's clause revision does not address this fundamental issue.

Some experts state bluntly that unless regulators introduce rules specifically targeting AI training data, corporate self-regulation through clause revisions alone cannot change the status quo of news website content being used without compensation.

Fair Use is a defense principle in U.S. copyright law that allows others to use copyrighted works to a limited extent without authorization, considering factors such as the purpose of use (commercial or transformative), the nature of the work, the proportion used, and the impact on the market. In the context of AI training, tech companies often argue that scraping data to train models constitutes "transformative use," but this claim is highly contested: whether the training process constitutes transformation and whether it harms the market for the original works—courts across different countries have yet to reach a unified conclusion. Meta's clause revision does not address this fundamental legal uncertainty, which is why experts believe contractual clauses alone cannot resolve the legality of AI training data.

The Dilemma Facing News Websites and Possible Ways Out

For news publishers, Meta's clause adjustment sends a complex signal: on one hand, the platform recognizes that public sensitivity to AI scraping is rising; on the other hand, the substantive boundaries of data use remain unclear. If publishers want to protect their content, they currently can only rely on technical measures (such as anti-scraping measures) and legal action, but both are costly and have uncertain effectiveness.

Some industry organizations are pushing for legislation that would require AI companies to obtain explicit authorization and pay fees before using news content. This direction aligns with the approach of the EU's Directive on Copyright in the Digital Single Market, but implementation in other jurisdictions still faces resistance. Meta's clause revision incident precisely highlights the enormous gap between corporate self-regulation and legal oversight.

The EU's Directive on Copyright in the Digital Single Market was adopted in 2019. Article 15 created a "neighboring right for press publishers," requiring online platforms to pay remuneration when using content from press publishers; Article 17 strengthened platforms' obligations to filter user-uploaded infringing content. The directive provides some legal leverage for news publishers, but its effect is limited to the EU, and the degree of transposition and implementation varies among member states. Other jurisdictions (such as the United States) lack similar specialized legislation, so industry organizations' efforts to push for legislation still face considerable resistance outside the EU.

Impact on the AI Industry

Meta is not the only company facing such questions. Companies like OpenAI and Google are also under scrutiny regarding their training data sources. If Meta's clause tweak is interpreted as "compliance theater," it could prompt other AI companies to follow suit—superficially revising terms while maintaining existing data acquisition strategies. This would further erode public trust in the AI industry's transparency.

From a broader perspective, the legal acquisition of AI training data has become a key bottleneck for industry development. Before clear regulatory rules are introduced, "technical adjustments" like Meta's clause revision will only recur, while the rights and interests of original content producers such as news websites remain unresolved.

One-line Summary

Meta revised its terms to remove AI training language but retained broad wording; legal experts say news sites' concerns remain unresolved.

Paragraph Summary

Meta recently revised its business terms, removing language that implied it could freely use news website content to train AI, but legal experts point out that this adjustment does not truly resolve publishers' concerns. The phrase "related to your product or service" remains in the terms and is extremely broad, potentially continuing to leave room for Meta to scrape news content. Experts question that the clause's ambiguity gives Meta excessive interpretive authority, and current law lacks clear definitions regarding AI training data scraping, with the fair use principle highly contested in the AI context. News publishers lack equivalent bargaining power and can only rely on costly and uncertain technical measures and legal action. Industry organizations are pushing for legislation requiring AI companies to obtain authorization and pay fees, but implementation faces resistance. Meta's clause revision incident highlights the enormous gap between corporate self-regulation and legal oversight; if interpreted as "compliance theater," it could prompt other AI companies to follow suit, further eroding public trust in the AI industry's transparency.