According to the US Copyright Office US Copyright, the generative outing of artificial intelligence based on text prompts – even detailed – is not protected by current copyright law.
The ministry has published these directives in a large report on policy issues concerning AI, focused on copyright of various IA outings. The document concludes that although generative AI can be a new technology, existing copyright principles may apply without changes to the law – and these principles offer limited protection for many types of work.
The new guidelines indicate that IA prompts are currently not offering enough control to “make users of an AI system the authors of the output”. (AI systems cannot contain copyright.) This is true if the prompt is extremely simple or involves long text chains and multiple iterations. “No matter how many times an prompt is revised and subject, the final release reflects the user acceptance of the interpretation of the AI system, rather than the paternity of the expression it contains,” said the report.
This decision apparently excluded protections for works such as “Space Opera”, an award -winning controversial image generated by priority, the creator of which fought an prolonged battle to record it with the Copyright Office.
The office demonstrates the unpredictability of AI systems with an image produced by the geminites of a cat smoking a pipe and reading a newspaper, noting that the Gemini ignored some quick instructions and added a few things apart – including a “In incongruous human hand”. He contrasted this process with the Jackson Pollock splashing method method, where he did not control the exact placement of painting on the web, but “controlled the choice of colors, the number of layers, the depth of texture , the placement of each addition to the composition set – and used its own body movements to execute each of these choices. “In the end, the office writes:” The question is the degree of human control, rather than the predictability of the result “.
“Regardless of the number of times an prompt is revised and subjected, the final release reflects the user acceptance of the interpretation of the AI system, rather than the paternity of the expression it contains.”
At the same time, Copyright Office says that using AI to help human creative production do not necessarily compromise the ability of this work to be protected by law. There is a difference between AI used as a tool to help creative work and “AI as a replacement for human creativity”, and the office says that more in -depth analysis is justified. But that ensures that using AI to describe a book or offer song ideas should not have an impact on copyright capacity of the final work produced by man, because the author is simply “referencing, but not incorporating, the exit”.
Artists can obtain some protection if they feed their own work in an AI system for modification – for example, using a tool to add 3D effects to an illustration. The elements generated by the AI of the work would still not be protected, but if the original product remains recognizable, “perceptible human expression” in the work could always be covered by copyright.
People can also receive protection for work that integrates the content generated by AI as long as there is an important creative modification. A comic strip with images generated by AI can be covered if a human organizes these images and associates them with text written on man, although the images generated by individual AI are not protected. Similarly, “a film that includes special effects generated by AI or background illustrations is copyable, even if AI’s effects and illustrations are not separately.” On a “determination on a case -by -case basis”, even the images generated by an invite could be protected if a human selects and remidms specific areas of the image. The office compares these scenarios to the manufacture of works derived from the copyright of art created by man – less the original human.
A separate question is whether the text invites themselves can be protected by copyright. Overall, the office has compared prompts to “instructions” that transmit not covered ideas, but it has recognized that some particularly creative people could include “expressive elements”. However, this does not translate into the work they produce by being protected.
The copyright office has not excluded the potential to change if technology is evolving. “In theory, AI systems could one day allow users to exercise as much control over how their expression is reflected in an output that the contribution of the system would become by heart or mechanical,” said the report. But for the moment, it does not seem to be “properly determine the expressive elements produced or control how the system translates them at the output”.
This document is part of a broader effort of the Copyright Office to clarify policy issues and identify legal gaps around AI, starting with a July 2024 report encouraging new Deepfake laws. The office then plans to publish a third and last report on its conclusions on “the legal implications for the training of AI models on works protected by copyright”.
