My personal opinion is that non-generative AI is not copyright-transitive the same way software is not copyright-transitive.
A Word document does not carry the copyright of Microsoft Word, because Microsoft Word is just a tool providing non-creative input to the document.
An AI model for face detection does not carry the copyright of the training data, because the training data is just a tool providing non-creative input to the model.
The training data *might* have been used in violation of some terms or license, but that is a *different* question to whether the resulting model itself is infringing on its copyright. Just like using a pirated version of Microsoft Word doesn't make the documents you make with it copyright violations.
It's GenAI where things change, because the output and the functionality of the model becomes (ostensibly) creative. And that's what copyright cares about, creative works.
Hoshino Lina (星乃リナ) 🩵 3D Yuri Wedding 2026!!!
npub1cx…z6wtz
2025-09-12 08:46:18 UTC
in reply to nevent1q…nfca
Author Public Key
npub1cxqje22r8ewnr30rt625d58cr8gpdjnw405cq5765m2qsvghnmpquz6wtzSeen on
wss://relay.momostr.pinkPublished at
2025-09-12 08:46:18 UTCEvent JSON
{
"id": "810ab43769bc5626fcbb39661775f50e3d7760b9667860de5e6025067c8aab83",
"pubkey": "c1812ca9433e5d31c5e35e9546d0f819d016ca6eabe98053daa6d40831179ec2",
"created_at": 1757666778,
"kind": 1,
"tags": [
[
"proxy",
"https://vt.social/@lina/115190450024101152",
"web"
],
[
"e",
"de73ed68101f82b8da1419e21a3965b103890cf325f15c6c6db88508f1c682c2",
"",
"root",
"daf360e6c6a1273c610269bee742cf99b70469de1d16a03b2fa89f00d1bb502e"
],
[
"p",
"daf360e6c6a1273c610269bee742cf99b70469de1d16a03b2fa89f00d1bb502e"
],
[
"proxy",
"https://vt.social/users/lina/statuses/115190450024101152",
"activitypub"
],
[
"L",
"pink.momostr"
],
[
"l",
"pink.momostr.activitypub:https://vt.social/users/lina/statuses/115190450024101152",
"pink.momostr"
],
[
"-"
]
],
"content": "My personal opinion is that non-generative AI is not copyright-transitive the same way software is not copyright-transitive. \n\nA Word document does not carry the copyright of Microsoft Word, because Microsoft Word is just a tool providing non-creative input to the document.\n\nAn AI model for face detection does not carry the copyright of the training data, because the training data is just a tool providing non-creative input to the model.\n\nThe training data *might* have been used in violation of some terms or license, but that is a *different* question to whether the resulting model itself is infringing on its copyright. Just like using a pirated version of Microsoft Word doesn't make the documents you make with it copyright violations.\n\nIt's GenAI where things change, because the output and the functionality of the model becomes (ostensibly) creative. And that's what copyright cares about, creative works.",
"sig": "bc626edc596e4bf033b66afa0d45ddf9e3c948cdf2f805409af97d1092d7b4f590bfff45adf0e29c0042b6f53684eafa6944ed6a872f02719970d572ce3c1928"
}