- Meta planned to introduce commercial terms allowing it to crawl external websites associated with business accounts to train its AI models.
- The terms, set to take effect on October 30, offered no opt-out mechanism for organizations using business tools or ad suites.
- Following severe pushback from news publishers and legal experts, Meta updated the language to remove explicit AI training rights.
The Data Scraping Controversy in Meta’s Business Terms
Tech giant Meta has revised its upcoming commercial terms of service following intense pushback from enterprise clients and news publishers. The original contractual language, scheduled to go into effect on October 30, gave the company broad permission to scrape and crawl external websites belonging to organizations that utilize its commercial services and feed that data into its AI models.
The policy applied across the board to any organization maintaining a commercial presence—including businesses utilizing Ads Manager, Meta Business Suite, WhatsApp AI agents, or verified enterprise pages. According to initial findings reported by Nieman Lab, virtually every online news publisher operates at least one active Meta business account, putting vast amounts of proprietary journalistic content within reach of the company’s automated web scrapers.
Automated Crawling and Legal Criticisms
The controversial clause authorized Meta and third parties acting on its behalf to deploy crawlers, bots, and automated user agents, even generating test credentials to access, index, cache, and retrieve updates from associated URLs. Customers had no way to opt out without ending their operational relationship with the social media giant entirely.
| Provision | Original Commercial Terms | Revised Post-Backlash Terms |
|---|---|---|
| Automated Site Crawling | Broad rights granted across all linked commercial domains | Crawling provisions partially rewritten |
| AI Model Training | Explicitly allowed to “train” and “develop” products | Word “train” and direct AI training clauses removed |
| Opt-out Choice | None (Hobson’s choice: accept or quit platform) | No explicit statutory robots.txt guarantee codified |
| Effective Date | October 30 | October 30 |
Danielle Coffey, president and CEO of the News/Media Alliance, condemned the move as an abusive exercise of market dominance, emphasizing that publishers could otherwise negotiate millions of dollars in formal content licensing agreements. Legal experts also raised alarms regarding contract enforceability. Eric Goldman, associate dean for research and professor of law at Santa Clara University School of Law, pointed out procedural flaws, describing the lack of a real opt-out mechanism as a classic Morton’s fork or Hobson’s choice, especially for clients with prepaid ad campaigns.
Meta Backpedals on AI Training Clauses
In response to mounting scrutiny, Meta updated its commercial terms preview late on Friday, officially stripping out the phrasing that permitted it to collect external site content to “train” its systems. While Gizmodo and other outlets confirmed the omission of the training term, several legal analysts and media watchdogs note that the company still retains substantial crawling leeway unless rigorous, explicit safeguards are cemented.
Enterprise accounts and publishing houses should verify which domains are directly linked to their Meta Business Suite and Ads accounts. Even though direct AI training language was struck, companies managing high-value intellectual property must maintain strict crawling directives and audit their robots.txt rules to safeguard their web assets.
Removing the explicit training language is an interesting pivot, but what happens to the automated crawling permissions that remain? Are they still indexing proprietary content under the hood, and how can publishers truly enforce robots.txt against embedded partner tags?