Meta has revised upcoming commercial terms after publishers and legal specialists questioned wording that appeared to give the company broad permission to collect material from websites associated with business accounts.
The revised language removes an explicit reference to training and narrows the stated uses to advertising and business products. However, the change does not remove the underlying permission for automated collection, according to Nieman Journalism Lab’s reporting on the revised document.
The distinction matters: a change in contractual wording is not evidence that Meta has stopped crawling websites or abandoned AI development. It also does not prove that any particular publisher’s material has already been collected under the new provision.
Why the original wording drew attention
Nieman Lab’s initial investigation examined terms scheduled to take effect on October 30. The concern involved websites connected with businesses using Meta’s services, extending the discussion beyond material uploaded directly to a social platform.
The original text described a range of automated activities, including collecting, indexing and revisiting website content. Legal specialists interviewed for that investigation questioned how broadly the resulting permission could be interpreted.
For publishers, the issue is the relationship between using a platform to reach an audience and granting rights over content hosted on their own websites. Those activities serve different purposes, even when a business links its website to its social accounts.
What changed and what did not
In its follow-up report, Nieman Lab said Meta had removed the word “train” and replaced a reference covering all Meta products with language focused on advertising and business products. The publication reported that lawyers still considered the remaining terms broad.
Their concerns are interpretations of the contract, not a court ruling. The revised wording’s practical and legal scope remains a separate question from the fact that the text was changed.
Meta has said the wording concerns its advertising and business services and does not change its web-crawling practices. The company’s explanation and the specialists’ concerns should both remain visible when describing the dispute.
Crawling, permission and training are different questions
Web crawling is the automated retrieval of material from websites. It can support functions such as discovering pages or obtaining information associated with a business. What an organisation may subsequently do with retrieved material depends on the relevant technical, contractual and legal circumstances.
AI training is a separate use of data. A reference to crawling does not, by itself, establish that a particular document entered a model’s training data. Likewise, deleting one word from a contract does not independently demonstrate how a company’s systems process information.
This is why precise reporting matters. A headline saying Meta has ended AI training would go beyond the development established here. The verified change concerns commercial terms and the debate about their scope.
The practical questions for website operators
The original investigation also examined the relationship between contractual permissions and robots.txt, a file through which website operators communicate access preferences to crawlers. Meta said owners could continue managing access through that mechanism, while the specialists questioned how the contract and those preferences would interact.
That disagreement remains important for publishers whose businesses depend both on distribution through social platforms and control over their own archives. A technical access instruction, a platform agreement and a content licence address related but distinct questions.
For any future assessment, the relevant evidence would include the final terms, Meta’s explanations and documented behaviour by its crawlers. Assertions that a specific paywall was bypassed or that particular articles trained a model would require evidence beyond the wording alone.
Frequently asked questions
Key questions and answers at a glance.
Did Meta stop all website scraping?
The reported revision does not establish that. Automated collection remains part of the dispute.
Did Meta announce an end to AI training?
No such conclusion follows from this change. Removing training language from these terms is a narrower development.
When are the commercial terms scheduled to take effect?
The reporting reviewed identifies October 30, 2026.
Are the lawyers’ concerns a court decision?
No. They are legal interpretations reported in the coverage. The draft does not present them as a settled judicial finding.
