With the rapid sophistication of generative AI technologies such as GPT-4 and Midjourney and the explosive growth in their creative uses, the U.S. Copyright Office announced today that when AI-generated material can be copyrighted. issued clarification guidance.
The guidance comes after the Copyright Office decided that authors could not copyright individual AI images used to illustrate comic books because each image was produced by Midjourney rather than a human artist. . In making that decision, the Copyright Office pledged to uphold a long-standing legal definition that the author of a creative work must be a human being to register the work. , confirmed that AI technology is not considered a creator.
It’s not the only thing that has influenced new guidance, but it’s the latest example. By tackling comics’ complex authorship issues, the Copyright Office will continue to explore broader copyright issues that arise as the AI models used to generate text, art, audio, and video continue to evolve. to launch an agency-wide initiative.
This guidance is analogous to the Copyright Office giving “instructions to commissioned artists” about what is not copyrightable for AI works generated solely by prompts without modification. I will provide some details. Since these works are not human works, they are not registered.
“When an AI technology receives only prompts from humans and generates complex written, visual, or musical compositions in response, the ‘traditional elements of the author’ are determined and determined by the technology, not the human user. will be executed,” explains the guidance. “Based on the Secretariat’s understanding of currently available generative AI technologies, users will have ultimate creative control over how such systems interpret prompts and generate materials. I will not exercise.”
However, as in Midjourney’s case, an author who arranges generative AI into a particular sequence (such as designing a comic book layout) may authorize that sequence of images if the arrangement is “creative enough”. rights may be granted. Similar logic applies if an author or artist modifies her AI-generated material and “that modification meets the criteria for copyright protection.” Examples might be modifying her AI image in Adobe Photoshop, or using a guitar pedal to modify his AI generated audio.
However, it is clear that the Copyright Office is in the early stages of navigating these complex cases, and the guidance remains somewhat vague. Authorities reviewing whether a person has been identified will make a decision on a case-by-case basis, the guidance said.
“The answer depends on the circumstances, especially how the AI tools behaved and were used to create the final work,” says Guidance.
All AI-generated content must be disclosed
Perhaps the most important aspect of the guidance is the author’s “obligation to disclose that works submitted for registration contain AI-generated content.”
Authors must distinguish between human-generated content and AI-generated content when registering their work. If the applicant is unsure how to reference her AI-generated content, the Copyright Office recommends providing a general statement that the work contains her AI-generated content. doing. This will encourage the Secretariat to follow up to assist each author in filling in the blanks on the application.
For artists with pending applications or who have already registered works containing AI-generated content, the Copyright Office suggests filing a supplemental registration to amend the publication record. . Failure to accurately reflect AI’s role in copyrighted works could result in “loss of the benefits of registration,” warns the Secretariat. This may make the work easier to copy with little or no legal recourse for claims of copyright infringement.
Failure to disclose AI-generated content is the only type of infringement covered by the guidance. Critics like Creative.ai co-founder Alex J. Champandard – a group of hackers and artists interested in generative AI –murmured It can be said that the current guidance puts the authors in a precarious catch-22 situation.
“By disclosing your AI, you are exposing yourself to compromise, but by not disclosing your AI, you are more secure, but you are being compromised. [the US Copyright Office]! ‘ Champagne d’art’s tweet suggested.
Ars was unable to contact Champandard to discuss other concerns his group had about the guidance.
