Meta has removed an explicit reference to artificial intelligence training from its revised commercial terms. Legal experts had warned that the original wording could allow the company to collect news articles and other website content without separate licensing agreements. The revision narrows the stated purposes for collecting that material, but questions remain about how broadly Meta could interpret its permissions.
The new terms are scheduled to take effect on Oct. 30, 2026. They allow Meta to use automated tools to access, analyze, collect and periodically retrieve content from websites associated with businesses’ products or services.
The issue matters to news organizations that use Facebook, Instagram, Messenger and WhatsApp to distribute journalism and reach audiences. Meta says the revision clarifies existing permissions and does not change how it uses web crawlers. However, lawyers who examined the wording say the remaining language could still encompass AI-related development.
Why legal experts remain concerned
The controversy began after the Nieman Journalism Lab reported on Oct. 15 that Meta’s original Commercial Terms explicitly allowed the company to use collected website content to develop, improve, research and train its products.
Three legal experts consulted by the publication warned that the wording could provide a basis for collecting publishers’ material without negotiating individual licensing agreements. The terms covered websites associated with business accounts, potentially extending their reach beyond posts published directly on Meta’s platforms.
Meta subsequently removed the explicit reference to training and limited the stated purposes to advertising and business products. However, the revised terms still permit the company to use collected content to develop those products.
Matt Topic, a media and intellectual property attorney at Loevy & Loevy, told Nieman Journalism Lab that the broad language concerning product development could still encompass AI products. James Grimmelmann, a professor of digital and information law at Cornell Law School, also questioned whether the revised wording sufficiently restricted the scope of the scraping permissions.
The terms authorize automated tools to access, index, cache, analyze, collect, retrieve and crawl websites. They also address the use of test credentials, which Grimmelmann said could potentially allow access to sites requiring registration. Meta has not publicly established that it intends to use this provision to access paywalled journalism.
Another concern involves robots.txt, a standard file website operators use to communicate instructions to automated crawlers. Meta says website owners can use the file to express their preferences. University of Wisconsin law professor Benjamin Sobel told Nieman Journalism Lab that such instructions do not necessarily create an enforceable legal restriction on a crawler.
The legal dispute centers on the difference between Meta’s stated purposes and the potential reach of its contractual permissions. The revised terms no longer explicitly mention AI training, but the lawyers’ objections concern whether the remaining language adequately limits what the company can do with collected material.
Facebook’s record of privacy disputes
Meta’s assurances come against a history of regulatory action involving Facebook’s handling of personal information and disclosures about data misuse. Those cases do not establish that the company is violating its revised terms, but they provide context for publishers assessing its commitments.
In 2012, the Federal Trade Commission (FTC) finalized a settlement resolving charges that Facebook had deceived consumers about their ability to control the privacy of their personal information. The order required the company to obtain users’ express consent before sharing information beyond their privacy settings and to maintain a comprehensive privacy program.
In 2019, the FTC imposed a US$5 billion penalty on Facebook to settle charges that it had violated the earlier order. The commission said the company had undermined users’ privacy choices, including through its handling of information shared with third-party applications.
The FTC also pursued a separate case against Cambridge Analytica, finding that the political consulting firm had used deceptive practices to collect personal information from tens of millions of Facebook users. That case concerned Cambridge Analytica’s conduct and should be distinguished from the FTC’s separate actions against Facebook.
The Securities and Exchange Commission (SEC) brought another case in 2019, alleging that Facebook had made misleading disclosures to investors about the risk of misuse of user data. Facebook agreed to pay US$100 million to settle the charges.
Together, these cases document a history of regulatory disputes over privacy representations, data handling and disclosure. They do not prove that Meta is violating the current commercial terms or intends to use scraped news articles to train AI systems. However, they help explain why publishers may seek clearer contractual guarantees rather than relying solely on public assurances.
What the terms mean for news publishers
The revised terms apply to businesses and organizations using Meta’s commercial products. For news publishers, the central concern is that the permissions cover websites associated with their businesses, not just articles and other material posted directly on Facebook or Instagram.
Meta’s language does not explicitly authorize AI training on all scraped content. The available evidence also does not establish that the company plans to use these permissions to avoid paying publishers for their work. Those are concerns about the possible reach of the terms, not established findings about Meta’s current practices.
Website owners can use robots.txt to communicate crawling preferences, according to Meta. Legal experts nevertheless question whether the terms provide sufficient contractual protection for those restrictions. The reported terms also do not offer businesses a partial opt-out that would allow them to continue using Meta’s commercial products while rejecting the scraping provision.
That leaves publishers weighing the value of Meta’s services for audience development against uncertainty about how the company may use material from their websites. The decision also raises a broader licensing question: whether publishers should have to accept broad collection permissions as a condition of using platforms that help distribute their journalism.
Before the terms take effect on Oct. 30, publishers need clearer answers on three points: whether scraped content can be used to develop or train AI products, whether Meta will honor website operators’ crawling restrictions, and whether businesses can reject those uses without losing access to its commercial services.