Unverified50% confidenceOpinionExact time
The Wikimedia Foundation emphasizes the need to distinguish between the "training" phase and the "output" phase of AI training: the training phase, in which language patterns are learned, does not constitute infringement, whereas the output phase, in which text is reproduced verbatim, may constitute direct infringement.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/30/2026
First Seen
Valid until: 1/28/2027
Sources
Related Entities
Related Claims
VerifiedLarge language models have two inherent weaknesses: their knowledge is cut off at training time and they cannot obtain the latest information, and they can only generate text and cannot directly perform actions.68% similarUnverifiedThe advantage of continued pre-training is that it can exploit massive unlabeled monolingual data, but the cost is that it requires a large amount of compute, and what is learned is only the distribution at the level of language form, not directly targeted at translation, the cross-lingual mapping task.68% similarUnverifiedCopyright lawsuits against language models typically focus on lyric text rather than audio recordings themselves, because text model training data is more likely to include lyrics scraped from the web67% similarUnverifiedWhen large language models are trained on text generated by other AI, the readability of the output declines.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1115065API
curl https://kongchang.com/api/v1/knowledge/claims/1115065MCP
get_claim(id=1115065)