In September 2025, Anthropic agreed to a settlement of 1.5 billion US dollars. The background was a class action lawsuit by authors who accused the AI company of using millions of copyrighted books without permission to train its language model Claude.
The case is considered one of the most important precedents to date in the dispute over AI training data and copyright.
WHAT WAS IT SPECIFICALLY ABOUT?
The lawsuit was filed in August 2024. The central allegations were:
Use of over 7 million books from unauthorized sources
Downloads from piracy websites
Additionally: purchasing, scanning, and subsequent destruction of physical books
Anthropic did not dispute AI training itself, but came under pressure due to the origin of the training data.
THE RULING: FAIR USE HAS CLEAR LIMITS
The presiding judge made clear:
✅ Training AI models with books can be Fair Use, as it is considered transformative.
❌ Obtaining data through piracy sources is not permissible – regardless of the purpose.
This was the first time a clear distinction was made between usage and data origin.
THE SETTLEMENT IN NUMBERS
Settlement amount: 1.5 billion USD
Affected works: approximately 500,000
Compensation: approx. 3,000 USD per work
Applicability: exclusively for past conduct
The settlement concludes the specific case but does not clarify all open legal questions for the future.
WHY THIS IS RELEVANT FOR COMPANIES
The case shows:
Legally uncertain training data is no longer a theoretical risk, but a massive liability and reputation problem.
For companies, this means:
Transparency of AI providers becomes crucial
Data origin and training methods must be traceable
"Black-box AI" is becoming increasingly risky – especially in Europe
WHY JURILO DELIBERATELY TAKES A DIFFERENT PATH
For precisely these reasons, Jurilo by Lawise.ai is hosted exclusively in Switzerland and Europe.
Our principles are clear:
No use of LLMs that were trained on allegedly stolen or unlicensed data
No training on user or customer data
Full alignment with Swiss and European legal standards (FADP, EU-compliant governance)
The Anthropic case confirms:
Legal certainty does not arise retroactively – it must be architecturally built in.
CONCLUSION
The 1.5 billion settlement marks a turning point for the AI industry.
It is not performance alone that determines the future of AI systems, but data origin, governance, and legal integrity.
For Europe and Switzerland, this is not a disadvantage – but a structural advantage.
