内容来源：https://www.wired.com/story/trump-administration-sides-with-ai-giants-new-york-times-lawsuit/

内容总结：

政府介入OpenAI与《纽约时报》版权之争，力挺AI“合理使用”

美国特朗普政府近日在OpenAI与《纽约时报》的法律纠纷中明确站队科技公司一方，引发广泛关注。2023年，《纽约时报》起诉OpenAI及其合作伙伴微软，指控其未经授权使用该报文章训练人工智能模型，侵犯版权。本周二，联邦政府向法院提交意见书，表明美国对AI训练使用版权内容是否属于“合理使用”具有“强烈利益关切”，因为这直接关系到美国AI产业能否“保持全球领导地位”。

政府律师在意见书中指出，《纽约时报》对“合理使用”的界定与现行版权法“不一致”，若该报胜诉恐将带来灾难性后果。意见书写道：“在误解合理使用原则的情况下限制大语言模型发展，将阻碍创新与科学进步，同时损害美国的繁荣与经济流动性。”政府强调，该意见不仅适用于本案双方，也对其他出版商和作者提起的相关诉讼具有参考意义。

政府认为，在版权作品上训练大语言模型具有“非同寻常的变革性”，即系统对输入内容进行了实质性改造，且OpenAI的模型并未与《纽约时报》文章形成显著竞争。法院在判定是否构成“合理使用”时通常会考量这些因素。意见书还警告，若将AI训练与其生成内容混为一谈，将对版权法整体产生“不利影响”，无异于认定人类作家通过研读他人作品来练习写作技艺也属违法。政府甚至举例类比，称此举如同少女时代的琼·狄第恩曾打字抄写海明威的小说以学习写作技巧。

司法影响与各方反应

尽管这份意见书不具强制约束力，但知识产权律师表示，主审法官几乎肯定会“认真对待”，因为它来自司法部，“天然具有很大分量”。加州大学伯克利分校法律与技术中心联合主任、版权律师帕梅拉·萨缪尔森称之为“重大进展”，与迄今AI版权案的判决方向一致。

《纽约时报》发言人则回应称，政府“站在少数市值万亿美元的AI公司一边，牺牲无数美国创作者的利益，而这些公司的产品正是建立在盗用创作者作品的基础之上”。他表示：“AI与创作者本可共同繁荣——AI公司只需按版权法要求，为支撑其产品的内容支付合理费用。”OpenAI和微软尚未回应置评请求。

其他原告方也对政府介入表示不满。作家协会首席执行官玛丽·拉森伯格称该意见书“充满错误论点，对合理使用原则和版权法存在严重误解”，该协会已于2023年对OpenAI提起独立诉讼。

背景：AI版权诉讼潮持续高涨

目前全美仍有数十起备受瞩目的诉讼，围绕AI公司使用版权作品训练工具的合法性展开。迄今为止的两大判决均发生在去年：Meta在卡德雷诉Meta案中技术上获胜，但法官指出原告未能充分证明损害，同时强调未经许可在特定情况下使用版权材料确实可能违法；而Anthropic则败诉，须向作者支付15亿美元赔偿，创下全国版权和解金额之最。该案法官特别裁定，AI训练本身属于合理使用，但因Anthropic盗版书籍仍需赔偿。

版权战火仍在蔓延。上周晚些时候，索尼和华纳音乐起诉Anthropic，此前环球音乐集团等巨头也已提起诉讼，Anthropic正以合理使用为由抗辩。各案均指控该公司盗用版权材料训练其热门AI工具Claude。

中文翻译：

特朗普政府刚刚在人工智能公司与《纽约时报》的法律纠纷中站到了OpenAI一边。

这家媒体巨头于2023年起诉了OpenAI和微软，指控这两家公司在未经许可的情况下利用其文章训练人工智能工具，违反了版权法。如今，政府介入其中，辩称OpenAI的行为并未越界。

周二，联邦政府向法院提交了一封信函，称美国对于“在受版权保护的作品上进行AI训练是否构成‘合理使用’”具有“强烈关切”，因为这关系到美国AI产业能否“保持人工智能领域的全球领先地位”。（“合理使用”规则保护对受版权保护材料的某些使用行为。）

信中称，《纽约时报》对合理使用的定义与现行版权法“不一致”，而该报胜诉可能带来灾难性后果。“在误解合理使用原则的情况下限制大语言模型的发展，将阻碍此类创造性及科学进步，同时损害美国的繁荣与经济流动性，”政府律师在信中写道。

虽然该信函聚焦于《纽约时报》与OpenAI，但政府指出，其论点适用于本案及相关案件中的所有当事方，包括其他出版商和作者。

政府认为，在受版权保护的作品上训练大语言模型被视为“具有非凡的变革性”，这意味着这些系统对其输入内容进行了有意义的改造。政府还写道，OpenAI的大语言模型并未与《纽约时报》的文章形成显著竞争。法院在判定是否满足合理使用原则时，会综合考虑这些条件以及其他因素。

信中称，将AI训练与人工智能工具的输出混为一谈，将“给版权法整体带来有问题的连锁影响”，因为这暗示人类作家通过研读他人文字来磨练写作技艺也同样是违法的。（律师们将此类比为少女时代的琼·狄第恩亲手打出欧内斯特·海明威的故事，以揣摩写作之道。）

Neal & McDevitt律师事务所的知识产权律师埃文·布朗告诉《连线》杂志，负责此案的美国地区法官西德尼·H·斯坦没有义务让这封信影响其最终裁决。然而，布朗表示，斯坦以及负责此类案件的其他法官几乎肯定会“非常认真地对待它，因为它来自司法部，本身分量极重。”

版权律师、伯克利法律与技术中心联合主任帕梅拉·萨缪尔森告诉《连线》杂志，政府的这份意见陈述是一个“重大进展”，并且与迄今的AI版权裁决方向一致。

“政府站在少数几家万亿美元级AI公司一边，牺牲的是无数作品被其窃取的美国创作者的利益。”《纽约时报》发言人格雷厄姆·詹姆斯告诉《连线》杂志。“AI和创作者本可以共同繁荣——AI公司只需按照版权法的要求，为其产品赖以存在的内容支付公平报酬即可。”

OpenAI和微软未立即回应《连线》杂志的置评请求。

其他原告对政府的介入表示不满。作家协会首席执行官玛丽·拉森伯格——该非营利组织于2023年对OpenAI提起了单独诉讼——告诉《连线》杂志，该组织对这封信“极为失望”，她称信中“充斥着错误论点，对合理使用原则和版权法存在严重误解”。

目前有数十起备受瞩目的诉讼正在进行中，涉及AI公司在受版权保护的作品上训练其工具是否合法的问题。迄今为止最大的两项裁决均发生在去年。Meta在Kadrey诉Meta案中技术上胜诉，尽管法官指出原告未能提供足够证据证明训练对其造成了损害，但法官同时强调，在其他情况下，未经许可在受版权保护的材料上进行训练确实可能违法。

与此同时，Anthropic败诉，必须向作者支付15亿美元的赔偿金，这是美国历史上最大的一笔版权和解金。该案法官做出了一个重要保留裁定，认定AI训练本身属于合理使用，但作者应获得赔偿，因为该AI巨头盗版了他们的书籍。

与此同时，围绕版权的更大规模斗争仍在继续。新的诉讼不断涌入；上周末，索尼和华纳音乐起诉了Anthropic，此前环球音乐集团等音乐巨头也提起了另一起诉讼，Anthropic正在该案中进行合理使用抗辩。在这两起案件中，这些公司均指控其受版权保护的材料被盗用，用于训练该公司广受欢迎的AI工具Claude。

英文来源：

The Trump Administration just took OpenAI’s side in the artificial intelligence company’s legal battle with the New York Times.
The media juggernaut sued OpenAI and Microsoft in 2023, alleging that the companies had violated copyright law by training their AI tools on its articles without permission. Now, the government has stepped in, arguing that OpenAI’s actions were within bounds.
On Tuesday, the federal government submitted a letter to the court stating that the United States has a “strong interest” in whether AI training on copyright works is considered “fair use,” because it connects to whether the country’s AI industry can “retain global leadership in artificial intelligence.” (The “fair use” rule shields certain uses of copyrighted materials.)
The letter argued that the Times’ definition of fair use is “inconsistent” with current copyright law, and that a win for the paper could have disastrous consequences. “Constraining LLM development under a misunderstanding of fair use doctrine would thwart such creative and scientific progress while hindering American prosperity and economic mobility,” lawyers for the government wrote.
While the letter focuses on the New York Times and OpenAI, the government noted that its arguments apply to all of the parties in this case and related cases, including other publishers and authors.
Training LLMs on copyrighted works is considered “extraordinarily transformative,” meaning that these systems meaningfully change their inputs, the administration argued. The administration also wrote that OpenAI’s LLMs do not significantly compete with The New York Times’ articles. Courts consider both of these conditions, as well as others, when deciding if the fair use doctrine is met.
Conflating AI training with what artificial intelligence tools produce would lead to “problematic implications for copyright law generally” by implying that it would be similarly illegal for human writers to practice their craft by training on the words of others, per the letter. (The lawyers compared it to how a teenage Joan Didion typed out Ernest Hemingway’s stories to figure out how to write.)
Evan Brown, an intellectual property lawyer with the firm Neal & McDevitt, tells WIRED that the US district judge overseeing this case, Sidney H. Stein, is not obliged to let this letter influence his ultimate decision. However, Brown says that Stein and other judges overseeing cases like this will almost certainly “take it quite seriously because it comes from the Department of Justice, and inherently carries a lot of weight.”
Pamela Samuelson, a copyright lawyer who codirects the Berkeley Center For Law & Technology, tells WIRED that the government’s brief is a “significant development,” and one which is consistent with AI copyright rulings thus far.
“The Administration is siding with a handful of trillion-dollar AI companies at the expense of the countless American creators whose work they stole,” New York Times spokesperson Graham James tells WIRED. “Both AI and creators can thrive–AI companies simply need to pay fairly for the content that makes their products possible, as copyright law requires.”
OpenAI and Microsoft did not immediately respond to WIRED’s request for comment.
Other plaintiffs are unhappy at the government’s incursion. Mary Rasenberger, the chief executive of the Author’s Guild—which filed its own lawsuit against OpenAI in 2023—tells WIRED that the nonprofit is “extremely disappointed” about the letter, which she calls ”replete with faulty arguments and a gross misunderstanding of the fair use doctrine and copyright law.”
There are dozens of ongoing, high-profile lawsuits over the legality of AI companies training their tools on copyrighted work. The two biggest decisions so far took place last year. Meta technically won its copyright case, Kadrey v. Meta, although the judge noted that the plaintiffs did not produce enough evidence that the training caused them harm while stressing that in other circumstances, training on copyrighted materials without permission might indeed be illegal.
Anthropic, meanwhile, lost its case, and must pay authors $1.5 billion in damages, the largest copyright settlement in national history. The judge in that case carved out an important caveat, ruling that the AI training itself was fair use, but that the authors were owed damages because the AI giant had pirated their books.
Meanwhile, the larger fight over copyright rages on. New lawsuits are still rolling in; late last week, Sony and Warner Music sued Anthropic, following another suit from other music giants like Universal Music Group, in which Anthropic is mounting a fair use defense. In both cases, the companies allege that their copyrighted material was stolen to train Claude, the company’s popular AI tool.
Comments
Back to top