Anthropic’s $1.5B Settlement and the New AI Training Data Liability

In late 2025, Anthropic settled Bartz v. Anthropic for $1.5 billion — the largest copyright settlement in the generative AI litigation wave. In January 2026, a $3.1 billion suit followed from Universal, Concord, and ABKCO alleging Anthropic trained Claude on torrented music. The two cases together establish a practical liability framework that every business using AI vendors needs to understand.

The legal framework, in short

Courts handling AI training data cases have drawn a clear distinction:

• Training on lawfully acquired copyrighted material: Likely fair use under existing doctrine. The Bartz court initially held this view.
• Acquiring training data by piracy: Not fair use. This is the liability hook in Bartz and the music publishers’ suit. The infringement isn’t the training — it’s the unauthorized acquisition.

What this means for AI vendors and their customers

If you are a business using a generative AI tool — for content creation, customer service, code generation, or document drafting — your exposure depends on how your vendor sourced its training data. If the vendor is later found to have trained on pirated content, that finding does not automatically transfer liability to you. But the disruption can be severe: model recalls, output restrictions, sudden price increases, or loss of indemnification.

Vendor due diligence questions to ask

• Provenance: How was the training data acquired? Licensed, scraped, or purchased? From whom?
• Indemnification: Does your contract include IP indemnification for the AI vendor’s training-data practices? What’s the scope and cap?
• Output insurance: If a court orders an AI system disabled or its outputs withdrawn, what’s your continuity plan?
• Audit rights: Can you request training-data attestations or audits? Most enterprise AI contracts now permit some form of this.

What this means for your business

AI vendor due diligence is no longer just a security and privacy exercise — it’s a copyright and continuity exercise. The Carlson Firm reviews AI vendor contracts for IP indemnification, audit rights, output continuity, and the data-provenance representations that determine whether you have recourse if your vendor ends up on the wrong side of the next Bartz-style settlement.