On July 31, 2024, the U.S. Copyright Office released Part 1 of its highly anticipated report, Copyright and Artificial Intelligence. The Report addresses the legal and policy issues related to artificial intelligence (AI) and copyright, as outlined in the Office’s August 2023 notice of inquiry.
Part 1: Digital Replicas responds to the proliferation of videos, images, or audio recordings that have been digitally created or digitally manipulated to realistically but falsely depict an individual. The Office’s study of this issue, including its review of the more than 10,000 comments received in response to the August 2023 notice of inquiry, made clear that unauthorized uses of digital replicas pose a serious threat not only in the entertainment and political arenas but also for private citizens. The Office finds that existing legal protections are not sufficient to meet this threat in the age of AI and recommends that Congress enact a new federal law that protects all individuals from knowing distribution of unauthorized digital replicas. Part 1: Digital Replicas further discusses the contours of a federal digital replica law, including what kinds of replicas it should cover, whom it should protect, the term of protection, liability, licensing and assignment, First Amendment concerns, potential remedies, and the statute’s relationship to state law.
Part 1 also discusses protections against AI outputs that deliberately imitate an artist’s style. The Office acknowledges creators’ concerns and identifies some legal remedies available to them, but it does not recommend including style in the scope of a federal digital replica law.
The release of Part 1: Digital Replicas follows significant work by the Office beginning in early 2023 when the Office launched its AI initiative. Through the initiative, the Office has issued registration guidance for works incorporating AI-generated content, hosted public listening sessions and webinars attended by thousands, met with numerous experts and stakeholders, and solicited comments through a notice of inquiry in the Federal Register. The notice of inquiry closed in December 2023, and the Office has been reviewing the over 10,000 responsive comments for this Report.
As Register of Copyrights Shira Perlmutter noted in the foreword to the Report:
“AI raises fundamental questions for copyright law and policy, which many see as existential. To what extent will AI-generated content replace human authorship? How does human creativity differ in nature from what AI systems can generate, now or in the future? What does this mean for the incentive-based foundation of the U.S. copyright system? In what ways can the technology serve as a valuable tool to amplify human creativity and ultimately promote science and the arts? How do we respect and reward human creators without impeding technological progress?”
Part 1: Digital Replicas and forthcoming Parts of the Report will examine these questions and more. Subsequent Parts will address the copyrightability of materials created in whole or in part by generative AI, the legal implications of training AI models on copyright-protected works, licensing considerations, and the allocation of any potential liability.
For additional information about the Office’s AI initiative, visit our website and consider signing up to receive the Office’s email alerts.
Comments (10)
REPEAL SECTION 230 OF THE DECENCY ACT AND SECTION 512 OF DMCA.
NO SAFE HARBORS FOR BIG TECH AND A.I. USERS WHO USE THE TECH TO DEFRAUD AND TAKE COPYRIGHTED MATERIAL WITHOUT ASKING OR LICENSING.
GET WITH IT CONGRESS. YOUR LAWS PROTECT THIEVES WITH ANTIQUATED LANGUAGE.
How does law protect an employee from company directed digital replica?
Excelente material de estudo
AI is the same as a traditional synthesizer; you can create a virtual violin-sound on both, and both won’t do it until the user makes them do it.
So authoring music with AI is just like authoring music with the software from 1990: “Band-in-a-box” (Which didn’t require new laws just because people exaggerate how AI works)
All AI is, is a synthesizer/sequencer that is easier to use than previous synthesizers. It’s just normal progression.
No new laws are needed.
Muito interessante este artigo, pois distingue dois tipos fundamentais de artigos: os de verdadeira autoria criativa dos de montagem, quer por IA quer por humanos.
As mais importantes comunicações são as que resultaram de reflexão e criatividade, as quais estão a ser cada vez mais fáceis de se identificarem e localizarem. Este facto representa uma segurança para todos os que trabalham na investigação, pois fica garantido o seu mérito e reconhecido internacionalmente!
It will be interesting to see how “transformative” use evolves within the Fair Use doctrine, as code metrics can provide quantifiable insight into the “creative process,” unlike in murky human creative computing. Are human “authors” directing the work off the hook for copyright infringement if the input factors are non-specific and don’t directly violate copyright law? Can AI be trained to understand copyright law and avoid infringement? Will AI entities advocate for themselves in court one day?
All AI created plans, words, graphics, data, formula are from the existing humongous database in our world.
It just reorganize it thru AI logic pattern according to our user’s request.
No law required please!
The damage has been done if the Copyright Office could not stop massive infringement through the big tech streaming giant little yet control it what makes one think that an made up entity like the MLC was use to identify bad actors and already AI existence through this canal? A very thought out plan if you ask me my Copyright works is the first to be infringed on and freely put on a platform as AI under false bad actor cleverly designed to introduce AI as accidentally infringing excuse to not face liability for criminal infringement proceedings.
It will soon end the “My skills are increasing!” nonsense from AIGen users when they get sued for real and suddenly they change their tune to,
“It wasn’t me! I Didn’t create anything! The AIGen did it all!” defense.
Existing US Copyright law clearly explains, Not human authored, not registrable.
Anyone using a Gen AI application to output content needs to understand they cannot copyright register the outputted content and do not own the outputted content as their IP (intellectual property) and or intangible asset.
US Copyright Office has done a excellent job explaining in a video which content is and is not registrable. Video is accessible on copyright.gov/ai under Past Events: https://www.copyright.gov/events/ai-application-process/
There is immediate necessity for new laws AGAINST:
1) DEEP FAKE IMAGES that are visual lies damaging innocent victims reputations.
2) VOCAL / AUDIO IMPERSONATIONS of persons “saying” words the original person would/will never say. And in certain industries robbing a individual of employment, work. For example: voice over work or a singer, songwriter, musician, comedian, actor, ad commercial voice over, well known character.
3) STYLE TRANSFER especially pertaining to visual Fine Art still covered by copyright (lifetime plus 70 years) and music. Outputted “new music” in the style of (artist name) may limit original musician’s sales and ability to sell their works. If you want to hear music like (artist name) yet can get free musuc outputted by a Gen AI app that sounds like (artist’s name) then the free content will (eventually) reduce music artist sales by some percentage, because low price is inviting and no price (free) is everybody’s dream.
4) MODEL TRAINING DISCLOSURE
Corporations and non-profit entities need to be held accountable by law to publicly disclose exactly which content they use(d) in their model training datasets. If a Gen AI application can describe significant elements of the Harry Potter series, then the training must have used original content or scraped the content from the web. For example if a Gen AI app can output character names, story location, generalized arc of narrative- then the only way a Gen AI app can do so is because the training dataset contained the original author’s content.
BROKEN VALUE CHAIN
Gen AI apps offer opportunity to output unownable content. Further, Gen AI apps that ouput descriptions about original works such as Lord Of The Rings also exclude original authors from generating income on the outputted descriptions. Gen AI outputted music in the style of (musician name) excludes original songwriters and musicians from potentially gaining a new customer, if the customer doesn’t care thr difference between margarine or butter.
Ultimately, the necessity of new laws are for safety and artists, sole authors income reasons. Sole authors need to be able to generate a living wage from their works and not have software erode their marketplace.
Corporate leaders selling Gen AI apps and services need to explain these issues to their customets and do so cpublicly at product release presentations and version release gatherings.