Meta's New Terms Spark Controversy: News Website Content May Be Used for AI Training for Free

Meta recently updated its platform terms of service, and the new wording may grant the tech giant the right to crawl and use third-party website content to train AI. According to Nieman Lab, critics worry that news publishers' content could unknowingly become free training data for Meta's AI models.

The Terms Change: From "Association" to "Training"

According to Nieman Lab, Meta's new terms will allow it to crawl and scrape content from websites "associated" with businesses that use Meta's platforms, explicitly including online publishers.

The new terms take effect on October 30. The text indicates that Meta can use this external content for "developing, improving, researching," and "training" — the last word is particularly critical, as it points directly to AI model training scenarios.

The core issue lies in the ambiguous boundary of the word "associated." If a news organization operates an official account on Facebook or Instagram, does its website constitute a site "associated" with Meta's platforms? If so, every article that organization publishes could legally be crawled by Meta and used to train its AI systems, without any payment or explicit authorization.

The Technical Meaning of "Scraping" and "Crawling"

The "crawling" and "scraping" mentioned in the article are two key steps in acquiring AI training data. Crawling typically refers to systematically traversing website pages and downloading content through automated programs (crawlers); scraping is more about extracting structured data from pages that have already been obtained. Together, they form the primary source of training corpora for large models. For news websites, their robots.txt file (a protocol file in the website's root directory used to tell crawlers which pages can be crawled) is usually the first line of defense, but robots.txt is not a legal document and relies only on crawlers' voluntary compliance. Meta's terms update effectively provides a legal authorization basis for crawling behavior, potentially allowing it to bypass purely technical conventions like robots.txt and convert what would otherwise require separate negotiation into blanket consent given at user registration.

The Dilemma Facing the News Industry

This terms change touches the most sensitive nerve in the news publishing industry. Over the past few years, news organizations worldwide have been grappling with tech platforms over content compensation.

From Australia's News Media Bargaining Code to the EU Copyright Directive, regulators have tried to force platforms to pay for the use of news content. However, Meta's terms update may circumvent these legal frameworks — because the terms define the use as "AI training" rather than traditional "content distribution" or "display."

For news organizations struggling to sustain their digital transformation, this is undoubtedly adding insult to injury. Their original reporting could be used to train AI models, and the content produced by those models could in turn divert their readers and advertising revenue.

The News Media Bargaining Code and the EU Copyright Directive

The Australian News Media Bargaining Code mentioned in the article, formally known as the News Media and Digital Platforms Mandatory Bargaining Code, took effect in 2021 and requires platforms such as Google and Meta to engage in mandatory commercial negotiations with Australian news organizations over the use of news content, with an arbitration body determining prices if no agreement can be reached. Article 15 of the EU Copyright Directive (adopted in 2019) created a "neighboring right for press publishers," granting publishers the right to authorize and charge for the online aggregation of their content. The core logic of both legal frameworks is to define the use of news content as "distribution" or "display" scenarios, thereby triggering payment obligations. Meta's new terms redefine the purpose of use as "AI training," which falls precisely into the gray area of existing legal frameworks — most jurisdictions have not yet issued final rulings on whether AI training constitutes "use" in the sense of copyright law, giving platforms an opportunity to argue that their conduct falls outside the scope of payment obligations.

The Gray Area of Law and Ethics

Meta is not the only tech company facing such controversy. In recent years, the copyright issue surrounding AI training data has become a global topic. Companies such as OpenAI and Google have all been sued for using publicly available web content to train models.

The cleverness of Meta's terms update lies in its attempt to establish new authorization rules for users and their associated websites through the service agreement. But the question is whether this kind of "blanket" consent is truly valid — when users "agree" to the terms in order to use the social platform, do they realize that their website content may be used for AI training?

From an ethical standpoint, even if the terms are technically legal, they are hard to call fair. Original content produced with resources invested by news organizations should not become free fuel for tech giants' AI arms race.

Key Cases in AI Training Data Copyright Litigation

The article mentions that companies such as OpenAI and Google have been sued for using publicly available web content to train models. The core dispute in these lawsuits is the applicable boundary of the "fair use" principle. In the United States, several representative cases are moving forward: for example, The New York Times v. OpenAI and Microsoft, alleging the unauthorized use of millions of articles to train GPT models; and a class action brought by multiple authors against Meta, alleging the use of copyrighted books to train the Llama model. The outcomes of these cases will directly affect the actual legal force of Meta's new terms — if courts ultimately determine that AI training does not constitute fair use, then even if authorization is stated in the terms of service, its legality may still be challenged; conversely, if the fair use defense succeeds, the binding force of the terms will be further strengthened. At present, there is no unified judicial consensus worldwide. Although the EU AI Act requires disclosure of training data summaries, it does not directly resolve the issue of copyright authorization.

Industry Reaction and Future Direction

Nieman Lab's report has drawn widespread attention from the news industry. It is foreseeable that around the time the terms take effect on October 30, more media organizations may speak out in opposition.

Some publishers may seek to withdraw from Meta's platforms, but given the traffic value brought by the massive user base of Facebook and Instagram, that decision is not easy to make. A more likely path is joint industry pressure demanding that Meta clarify the boundaries of the terms, or legal challenges to their validity.

The legality and reasonableness of acquiring AI training data are becoming the focus of a new round of contention between tech companies and content producers. Meta's terms update is just the latest act in this long-running battle.

One-line Summary

Meta's new terms may allow it to scrape associated website content to train AI, sparking strong concerns in the news industry.

Paragraph Summary

Meta updated its platform terms of service, and the new wording may allow it to crawl third-party website content "associated" with Meta's platforms for AI training, explicitly including news publishers. The new terms take effect on October 30, and the core dispute is the ambiguous boundary of the word "associated" — if a news organization operates an account on Facebook or Instagram, does its website constitute an "associated" site? Critics worry that original reporting will become free training data for Meta's AI models without knowledge or compensation. This move may circumvent existing payment frameworks such as Australia's News Media Bargaining Code and the EU Copyright Directive, because the purpose of use is defined as "AI training" rather than content distribution. The article also explores the legal and ethical gray areas, industry reaction, and future direction.