Gather — 3 July 2026 · Zeitgeist — a spike by Chris Gathercole

Like Company v Google CJEU Holds First-Ever Hearing on Generative AI and Copyright (Bird & Bird, 2026) — On March 10, 2026, the CJEU Grand Chamber (15 judges) held the first hearing directly asking whether training a large language model violates EU copyright law. The case is Like Company v. Google Ireland Limited (C-250/25). The Advocate General’s opinion is due September 3, 2026 — the first formal legal milestone in the EU’s determination of whether commercial LLM training falls within the text and data mining (TDM) exception.
EU copyright law roundup — first trimester of 2026 (Kluwer Copyright Blog, 2026) — Covers all EU-level copyright developments in Q1 2026 including the Like v. Google hearing, the European Parliament resolution on copyright and generative AI (March 10, same day), and the interaction between the AI Act transparency requirements and existing copyright law.

Court Rules AI Training on Copyrighted Works Is Not Fair Use (Davis+Gilbert LLP, 2026) — The judge found that ingesting entire copyrighted works for commercial model training goes well beyond fair use — the finding that underpinned Anthropic’s $1.5B authors’ settlement. This is the most concrete US judicial statement yet on the limits of fair use for LLM training, and it will be cited in every subsequent case.
GEMA v. OpenAI — Europe’s Direction on AI Infringement? (William Fry, 2026) — Analysis of the November 2025 Munich Regional Court ruling (first European court to directly address AI training copyright): ChatGPT unlawfully used copyrighted German song lyrics. Expected at Munich Court of Appeal in 2026. The Munich reasoning parallels the US fair use rejection — convergence across jurisdictions.

AI Copyright Lawsuit Developments 2025: A Year in Review (Copyright Alliance, 2026) — Five major publishers (Hachette, Macmillan, McGraw Hill, Elsevier, Cengage) filed a lawsuit on May 5, 2026. The publisher wave follows the settlement by music labels, author class actions, and news publishers — the litigation front is expanding across creative industries simultaneously.

AI Training in 2026: Anchoring Synthetic Data in Human Truth (Invisible Tech, 2026) — At current scaling rates, frontier labs will exhaust all available public web text by 2028. Licensing deals cover a fraction of what the models actually need. Synthetic data fills the gap — but the central challenge is preventing “model collapse” (training on AI-generated output that lacks ground-truth signal). The article argues for “anchoring” synthetic data in human-validated ground truth as the structural safeguard.
The price of AI training data, from $5M to $250M (Quartz, 2026) — Real deal structures emerging: Amazon/NYT deal $20–25M/year; News Corp average $50M/year across WSJ, New York Post, global titles. Synthetic data market valued at $2.1B in 2025, growing at 35.2% CAGR. The gap between licensed and needed data is accelerating both market prices and synthetic data investment.

Open vs Closed AI Ecosystems The CJEU hearing on TDM exception directly affects open-weight model development — if commercial training without opt-out is ruled out, open-weight models in the EU face the same constraints as closed ones.
AI Impact on Society The Fable 5 export control episode adds a new dimension: government can control frontier model access through export controls, not just copyright law. Two distinct regulatory levers now exist.

Emerging theme: The CJEU Advocate General opinion (September 3, 2026) is the single most consequential upcoming legal event in this space — it will shape whether the EU TDM exception covers commercial LLM training.
Quality signal: Davis+Gilbert’s summary of the fair use ruling is the clearest authoritative statement yet — bookmark as the primary reference for the US fair use position on training data.
Keyword suggestion: "TDM exception" LLM EU OR CJEU OR "Like Company" to track the September AG opinion and its aftermath.