Kleiner v. Adobe is another step in a pattern: the legal system is increasingly treating “training data governance” as a compliance domain, not a research footnote.

2025-11-28 · Source: Pascal’s Substack · Field: Legal & Regulatory — Intellectual Property & Patents, Compliance & Risk Management, Litigation & Dispute Resolution · Depth: Advanced, medium

Summary

The "Kleiner v. Adobe SlimLM" lawsuit, filed February 9, 2026, in the Northern District of California, is a proposed class action alleging Adobe Inc. trained its SlimLM small language models on large-scale, unlicensed copies of copyrighted books, including author Arthur Kleiner's registered work. The complaint asserts a "dataset supply chain" infringement theory, claiming SlimLM was trained on SlimPajama-627B, a dataset derived from RedPajama, which allegedly incorporated "Books3"—a corpus associated with pirated books from shadow libraries. Kleiner's suit alleges direct copyright infringement under 17 U.S.C. § 501, seeking damages, attorneys' fees, injunctive relief, and destruction of infringing copies under 17 U.S.C. § 503(b). The case also highlights Adobe's "ethical AI" marketing against its alleged reliance on known unlicensed datasets.

Key takeaway

For CTOs and VPs of Engineering integrating language models, this lawsuit signals that you cannot outsource legal risk to open dataset supply chains. Your teams must implement robust data governance and provenance tracking for all training data, especially for commercialized models. Proactively verifying the licensing and origin of datasets like Books3 is crucial to mitigate significant copyright infringement liability and reputational damage.

Key insights

AI training data governance is evolving into a critical legal compliance domain, not merely a research consideration.

Principles

• Dataset supply chain liability is emerging.
• Fair use defense is fact-intensive.
• Piracy-tainted sources increase litigation risk.

In practice

• Audit dataset provenance rigorously.
• Document internal data governance.
• Filter out known pirated content.

Topics

• AI Training Data
• Copyright Infringement
• Fair Use
• Data Governance
• Small Language Models

Best for: CTO, VP of Engineering/Data, Executive, Legal Professional, AI Ethicist, Director of AI/ML

Related on AIssential

• The legal theory used against commercial AI companies may also reach academic AI research, open models, university labs and public-interest research infrastructure. — The Hendrix v. Apple case could redefine fair use for AI training, impacting academic research and open models.
• Penguin v. OpenAI: AI-assisted production of market substitutes in the very formats that already plague children’s publishing (fake titles, lookalike covers, rapid self-publishing). — AI models reproducing copyrighted content on demand can constitute actionable copyright infringement.
• Intellectual Property vs. Machine Learning: Data Ownership and AI Models in the Age of Litigation — AI's reliance on vast datasets without clear consent creates profound, unresolved intellectual property and ownership disputes.
• Authors' lucky break in court may help class action over Meta torrenting — The *Cox v. Sony* Supreme Court ruling on ISP liability for piracy is now central to Meta's AI training data copyright defense.
• Encyclopedia Britannica sues OpenAI for training on nearly 100,000 articles without permission — AI models' ability to reproduce copyrighted content from training data is a central legal and technical challenge.
• Microsoft trained its MAI models on unlicensed web data despite promising "enterprise grade, clean and commercially licensed data" — Microsoft's MAI models used unlicensed web data, challenging its "clean data" claims amidst ongoing fair use legal disputes.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Pascal’s Substack.