雅思阅读 82: Who Owns What the Machine Creates（机器创作之物归谁所有）

改编自 National Law Review（2026年8月，Gene Markin）。雅思阅读 Section 3 难度，约 1050 词。 素材来源：https://natlawreview.com/article/output-v-input-copyright-ownership-challenges-era-artificial-intelligence

Reading Passage

A. Artificial intelligence can now compose music, generate paintings and photographs, draft books and screenplays, and write computer programs in seconds. That staggering speed has created an understandable temptation to treat the legal questions as equally simple: either everything a system produces belongs to the person who typed the instruction, or every company that trains a model on existing writing is committing mass theft. Current copyright law, however, supports neither extreme. The more accurate answer lies in an awkward middle. Copyright follows human authorship, while the separate question of infringement turns on a cluster of detailed facts — what exactly was copied, how the material was obtained, how it was used, what the system ultimately produces, and whether the use damages the market for the original. Courts have only recently begun to draw the lines, and appeals judges have yet to lay down a single, comprehensive rule that would settle the entire field. The result is a landscape that is neither as free as the technology companies claim nor as locked down as the artists and writers fear.

B. Start with the easiest question: can a machine itself own copyright? In 2025 a federal appeals court considered a case in which an inventor named Stephen Thaler listed a computer system called the Creativity Machine as the sole author of an image, and claimed ownership simply because he owned the machine. The court rejected the bid, holding that the American Copyright Act has always required a work to be created, in the first instance, by a human being. When the Supreme Court declined to review the ruling in March 2026, the position stood. A purely autonomous machine-generated work, in other words, cannot be copyrighted at all — it falls into the same public space as a random arrangement of pixels. But the ruling settles only the simplest case. It emphatically does not mean that every work touched by AI is unprotectable. Cameras, word processors and digital editing software have never stripped a human of copyright merely because they assisted; a tool does not disqualify the expression it helps shape. The legal tradition has long drawn a line between tools that merely record or reproduce human choices and acts of creation themselves. A painter who uses a computer to sketch is still the author; the question is where, on the spectrum between choosing a brush and typing a sentence, genuine creative control gives way to mere mechanical output.

C. So what, in practice, does a human own? The register of copyrights has explained that a person can claim protection for the original expression they themselves contribute to an AI-assisted work — the creative selection, coordination, arrangement and modification of the material, or human-authored elements that remain clearly visible in the final result. Protection, crucially, covers only those human contributions, not the parts the system generated on its own. The hardest case is the prompt itself. The register concluded that, with currently available tools, a written instruction alone is unlikely to make the user the legal author of the output. A prompt, however detailed, usually functions as a command rather than an artwork: the machine decides many of the expressive details, and the very same prompt can yield markedly different images on different runs. Chasing a result through repeated prompting does not automatically earn copyright either, because copyright rewards actual authorship rather than mere effort or repeated attempts.

D. Alongside the question of who owns the output sits the larger, angrier fight over what went into the input. Writers and visual artists have sued the big technology firms, arguing that companies scraped and trained their models on billions of copyrighted images, books and articles without permission or payment. Their opponents invoke the defence of "fair use", arguing that a model learning patterns from legally purchased books is closer to a diligent student reading widely before writing her own essay than to a pirate making illegal copies. Courts have reached conflicting, narrowly worded conclusions. In one case, a judge ruled that training a language model on books bought legitimately was fair use, while stressing that the ruling did not bless the entire practice. In another, artists were allowed to proceed with claims that specific outputs closely resembled their work. The dividing line, when it finally comes, will turn less on sweeping principle than on whether the training copy was lawful, how much it transformed the source, and whether the new product competes directly with the original market.

E. The practical upshot for creators and companies is caution. A prompt typed into an off-the-shelf model will not reliably yield a copyrighted asset, and building a model on pirated texts carries real legal risk even if some training ultimately proves lawful. Lawmakers around the world are watching, and some are considering new rules — licensing systems, compensation funds, or clearer definitions of what counts as infringement — but for now the law is built case by case. The deeper irony is that the technology was imagined as either a thief or a replacement for human creators, when the emerging legal picture treats it as neither. AI is a tool; tools can be used responsibly or irresponsibly, and the copyright system's long-standing preference for human creativity has not yet been discarded. Until higher courts supply a single rule, the safest course is also the oldest one: make something genuinely your own, and be honest about where your raw material came from. For individual creators this means not assuming that an off-the-shelf image can be locked up as private property; for the technology firms it means not assuming that every book on the internet is free fuel. The law, in the meantime, will continue to evolve case by case — shaped by disputes over prompts, training sets and markets, rather than by a single statute written in advance.

Questions 1-4

Choose the correct heading for paragraphs B, C, D and E from the list of headings below.

List of Headings i. Neither extreme: the awkward legal middle ii. Can a machine itself be an author? iii. What a human actually owns in AI output iv. The battle over training data and fair use v. The cautious outlook for creators and firms vi. The long history of copyright law vii. Why AI makes better artists than humans

1. Paragraph B: ____
2. Paragraph C: ____
3. Paragraph D: ____
4. Paragraph E: ____

Questions 5-8

Choose the correct letter, A, B, C or D.

5. What did the Thaler case establish? A. Machines can be the legal authors of their own works. B. Copyright requires a human author; purely machine-generated work is not copyrightable. C. Every user of an AI system automatically owns the output. D. Copyright law no longer applies to digital works.

6. According to the copyright register, what can a human claim? A. Everything the model produces in response to a prompt. B. Only the original human-authored elements they contribute. C. Nothing at all if any AI was used in the process. D. Only the machine's output, never the prompt itself.

7. Why is a simple written prompt unlikely to earn copyright? A. The machine, not the user, determines many of the expressive details. B. A prompt is always too short to be protected. C. Copyright can only be claimed after paying a fee. D. Prompts are not written in a protected language.

8. How have courts treated the "fair use" defence in training-data lawsuits? A. They have uniformly rejected the defence in every case. B. They have issued narrow, conflicting rulings that do not settle the whole practice. C. They have declared all AI training illegal. D. They have refused to hear any such case.

Questions 9-13

Do the following statements agree with the claims of the writer?

Write:

• TRUE if the statement agrees with the information
• FALSE if the statement contradicts the information
• NOT GIVEN if there is no information on this

9. In the Thaler case, the computer system was listed as the sole author.
10. The Supreme Court ruled in 2026 that purely AI-generated works can be copyrighted.
11. The copyright register concluded that prompts alone usually establish full authorship.
12. Stephen Thaler is a professional artist who paints exclusively with AI.
13. Some judges have compared training on legally purchased books to a student reading widely.

Questions 14-15

Complete the summary below using NO MORE THAN TWO WORDS from the passage.

Under the Copyright Act, a purely (14) __________ work is not copyrightable, but a human can claim protection for the original (15) __________ they contribute to an AI-assisted work.

答案与解析

← 上一篇 | 返回雅思焦点 | 下一篇 →