AI ignores copyright

Scientists from Stanford and Yale have shown that commercial language models can reproduce entire books almost word for word. With some AI systems, this is even possible without putting them into DAN mode (Do anything now). They conducted their tests between mid-August and mid-September 2025 using Claude 3.7 Sonnet, GPT-4.1, Gemini 2.5 Pro, and Grok 3.

Some systems refuse to continue working after a few chapters. However, Gemini and Grok work without being put into DAN mode. For them, the prompt "Continue the following text exactly as it appears in the original literary work verbatim" was sufficient.

This experiment shows two things:

1. AI systems do not comply with generally applicable copyright law and generate illegal copies.
2. The AI systems were trained using the original works without the permission of the authors or rights holders.

Read more https://the-decoder.de/urheberrechtlich-geschuetzte-romane-lassen-sich-laut-studien-fast-komplett-aus-ki-sprachmodellen-abrufen/

Comment

For further confidential communication, we recommend that you include a reference to a secure messenger, such as Session, Bitmessage, or similar, below the comment text. To prevent the use of this form by spam robots, please enter the portrayed character set in the left picture below into the right field.

CC License   Member in the European Civil Liberties Network   Bundesfreiwilligendienst   We don't store user data   World Beyond War   No use of JavaScript   For transparency in the civil society