Back to News
PolicyAI Understanding briefing

US appeals court backs Thomson Reuters in AI copyright case

The 3rd US Circuit Court of Appeals affirmed a lower‑court ruling that ROSS Intelligence infringed Thomson Reuters’ Westlaw headnotes, marking the first appellate decision on AI training fair‑use claims.

4 min readRead the linked source
Source-provided image accompanying US appeals court backs Thomson Reuters in AI copyright case
Source referenceSource recorded
Publisher
finance.biggo.com
Source link
finance.biggo.comhttps://finance.biggo.com/news/8db00528-9ad9-4743-a2c4-0e3deb23c5c9
Source type
Linked source — primary-source status has not been established.
ContextUnderstand this in 60 seconds

Start here

Key terms

Generative AI
AI systems that produce new content such as text, images, audio, video, or code.
Test yourselfAI Ethics Quiz

What happened

The 3rd US Circuit Court of Appeals rejected ROSS Intelligence’s fair‑use defense, upholding Thomson Reuters’ copyright claim over the use of Westlaw headnotes to train a legal‑search AI.

On Tuesday, the Philadelphia‑based 3rd US Circuit Court of Appeals affirmed a district‑court finding that ROSS Intelligence copied roughly 2,243 Westlaw "headnotes" – concise legal summaries – to build its AI‑powered legal‑research platform. The appellate court rejected ROSS’s argument that its use qualified as fair use, noting that the startup used the material to create a competing search tool rather than a transformative generative model.

The court’s detailed reasoning remains sealed, but the outcome aligns with Judge Stephanos Bibas’s 2025 district‑court opinion that ROSS’s use was not transformative. ROSS had previously sought a licensing agreement with Thomson Reuters, was denied, and instead obtained about 25,000 "Bulk Memos" from a third‑party provider derived from the same headnotes.

The ruling is the first appellate decision directly addressing AI training under U.S. copyright law. While it applies to a non‑generative legal search engine, experts caution that it does not automatically extend to large language models such as ChatGPT or Gemini, which may argue a different transformative use.

Source details: finance.biggo.com ↗

Why it matters

The decision sets a precedent for how U.S. courts may treat the use of copyrighted material in AI training, signaling that non‑transformative, data‑driven AI tools could face liability and that licensing costs may rise.

The case provides an early appellate precedent for the many pending lawsuits where publishers allege that AI developers copy copyrighted works for training. If courts adopt a narrow view of fair use, AI firms could be required to negotiate costly licenses for high‑quality proprietary data, favoring well‑capitalized companies.

The decision underscores the importance of the four fair‑use factors—purpose, nature, amount, and market effect—highlighting that non‑transformative, competitive uses are vulnerable. This could accelerate the formation of voluntary licensing markets, a stance the U.S. Copyright Office has already expressed support for.

Internationally, the ruling contrasts with the European Union’s approach, which is moving toward a registry that lets rights holders opt out of text‑and‑data‑mining. The divergent regulatory philosophies may affect where AI developers source data and how they structure their compliance strategies.

Interactive Mechanism

Interactive Mechanism: How It Actually Works

Explore the underlying technology behind this development interactively.

Thinking Budget (Test-Time Tokens):1,024 tokens
Complex Accuracy79%Math & Code Logic
Latency3.2sTime to first full output
Inference Cost$0.0092Per query estimated
Reasoning StyleStep VerificationInternal chain depth
Active Thinking Trace:
1Deconstruct user problem into formal constraints
2Propose candidate hypotheses & step-by-step calculation
3Self-correction: Backtrack and refute subtle edge cases
4Exhaustive consistency check & final output synthesis
Core takeaway: Test-time compute fundamentally changes AI economics. Instead of only scaling during pre-training, giving reasoning models more tokens at inference time allows them to systematically solve PhD-level STEM problems.
Interactive Concept Check+10 Points
AI Ethics Quiz

Impossibility results in algorithmic fairness (e.g. Kleinberg et al., Chouldechova) show what?

What to watch next

Future litigation over training data, potential legislative moves on AI‑specific copyright rules, and the development of voluntary licensing markets.

How lower courts apply the appellate reasoning to models, especially in high‑profile cases involving OpenAI, Microsoft, and other large developers.

Potential legislative proposals in Congress that could codify licensing requirements or create a compulsory data‑licensing framework for AI training.

The evolution of voluntary licensing initiatives, such as the Copyright Office’s proposed marketplace, and whether they can scale to meet the data needs of large AI models.

Related guides & quizzes

AI EthicsAI TrainingFuture of AITest what you know — try a free AI quizLook up an AI term in our glossaryFollow the AI regulation tracker
Found this useful?