怎么查 AI 生成内容的版权风险How to Check Copyright Risk for AI-Generated Content

输入侧查来源：中国办法要求「使用具有合法来源的数据和基础模型」。

输入侧查例外：欧盟《数字单一市场版权指令》为文本与数据挖掘设了例外，科研与商业口径不同。

输出侧别默认归属：AI 生成内容能不能受版权保护，各法域结论不同，先查当地口径。

商用前读厂商条款：模型使用条款对商用、竞品、输出归属各有规定。

留来源记录：素材从哪来、训练数据什么构成，纠纷时来源链第一。

诉讼还在进行：判例会变，重大商用决策找法务，别照新闻下结论。

Input side, check provenance: Chinese rules require "lawfully sourced data and foundation models".

Input side, check exceptions: the EU Directive on Copyright in the Digital Single Market sets text-and-data-mining exceptions, with research and commercial terms differing.

Output side, do not assume ownership: whether AI-generated content can be copyrighted differs by jurisdiction; check local practice first.

Before commercial use, read the vendor terms: model terms of use differ on commercial use, competing products, and output ownership.

Keep provenance records: where materials came from and what the training data was — the chain of provenance comes first in disputes.

Litigation is ongoing: case law changes; for major commercial decisions consult legal counsel, not news headlines.