AI Governance Institute
← News
Corporate Policy2026-10-23

Meta's Revised Terms Still Permit AI Training, Legal Experts Warn

What happened

Meta revised its business account terms to drop explicit language about using scraped web content to 'train' AI models, a change that followed criticism from news publishers. The update was analyzed in a Nieman Lab report via The Verge published October 22, 2026. Cornell Law professor James Grimmelmann and media attorney Matt Topic both told Nieman Lab that the word 'develop' is wide enough to encompass model training. They noted this word remains in the revised terms. They also said robots.txt signals are not contractually binding under the new terms. Publishers use these signals to tell web crawlers not to copy their content. The signals may provide no legal protection. Meta maintained that it continues to honor robots.txt preferences, but the outside legal analysis directly contradicts the idea that publishers have gained meaningful protection. This follows earlier coverage of Meta's October 30 Terms Shift Training Data Liability to Publishers, which addressed a related but distinct terms change event.

Why it matters

  • ·Organizations that publish web content may have no contractual basis for assuming robots.txt signals or revised platform terms protect them from AI training use. This increases their exposure to data-use disputes.
  • ·Compliance teams that reviewed Meta's earlier terms change as a sufficient safeguard may need to reopen that assessment. Expert analysis suggests the revised language does not materially narrow Meta's rights to use scraped content.
  • ·The gap between a vendor's public announcement and what its terms actually permit is a vendor due diligence failure point. Legal teams should not treat press-release-level assurances as a substitute for independent review of the actual contract language.

Governance controls affected

DGC-001Training Data ProvenancePRC-001AI Vendor Due DiligencePRC-002AI Contractual RequirementsPRC-003Third-Party AI Model EvaluationPRC-006Vendor Safety Commitment Verification

What to do now

  • ☐Have your legal team review Meta's current business account terms directly, focusing on whether the word 'develop' or similar broad language could cover AI model training under applicable law.
  • ☐Ask your legal and procurement teams whether any existing agreements with Meta or similar platforms include explicit, enforceable prohibitions on using your content for AI training, rather than relying on robots.txt signals.
  • ☐Audit which of your organization's web-published content, including news feeds, articles, or data, may be accessible to Meta's crawlers, and assess whether that exposure is acceptable given the unresolved legal questions.
  • ☐Update your vendor due diligence process to require independent legal review of terms changes at major AI platform vendors, not just acceptance of the vendor's own characterization of what changed.
  • ☐If your organization has content licensing agreements or is considering them, ask counsel whether those agreements would override broad platform terms of the kind described here.

What to watch next

Courts have not yet resolved whether broad 'develop' language in platform terms authorizes AI training on scraped content, and litigation involving scraped training data is active across multiple jurisdictions. Compliance teams should monitor clarifying rulings or enforcement actions under the FTC Enforcement on AI (Section 5 of the FTC Act). They should also watch any legislative movement that would give robots.txt signals statutory force. Publisher trade groups are pressing for explicit licensing requirements, and any shift in that direction would require rapid contract and terms review across organizations with significant web-published content.

Stay ahead of stories like this

Get every US AI governance development like this one, plus the rest of the week's developments. Every Thursday.

Free every Thursday. Unsubscribe anytime.

Related Coverage

Corporate Policy2026-10-16

Meta's October 30 Terms Shift Training Data Liability to Publishers

Meta updated its commercial terms, effective October 30, 2026, to authorize automated crawling of websites linked to business accounts and the use of that content for AI training. Legal experts warn the terms may allow Meta to argue they override robots.txt blocking instructions and existing content licensing agreements. Enterprise compliance teams managing AI training data provenance and publisher licensing programs face direct exposure.

Research2026-10-13

Startup Graveyard Data Is Flowing Into AI Training Without Worker Consent

AI labs are purchasing operational data archives from failed startups, brokered by wind-down services such as SimpleClosure and Sunset. The archives contain employee emails, internal chat histories, calendars, and code repositories. No federal law currently prohibits these sales, leaving workers with no notice, no opt-out, and no recourse.

Research2026-10-17

90% of US Work Hours Lack AI Benchmark Coverage, Mercor Research Finds

Mercor Research mapped 29 AI benchmarks against nearly 19,000 job tasks across 923 US occupations. The study found that only 208 occupations have meaningful benchmark coverage. This means vendor capability claims based on benchmarks may not reflect how AI actually performs on most real workplace tasks.