Meet Maya AI Logo

AI and copyright infringement

Meet Maya AI AI, copywriting, and infringement

This article explores how artificial intelligence-generated works can be protected under current copyright legislation. It defines AI and distinguishes it from sentient robots, which gives a crash course in US copyright law, and describes how AI is expected to replace human employment.

It outlines how copyright protection will be given to either the principal of the AI or the AIs end-user, the legal implications of recognizing AI as a corporation, and the public domain status of AI-produced works. 

It also looks at non-legal concerns, including paying the AI agent, the impact of replacing human workers, and using insurance to make up for AI copyright infringement.

AI is a set of methods or instructions that allow machines to mimic some aspects of organic cognition. Machine learning is the foundation for nearly all modern AI techniques.

Table of Contents

US Copyright Office Regulations

The US Copyright Office (USCO) has rejected Stephen Thaler’s copyright request for an AI-generated work of art. In November 2018, Thaler visited the USCO with the illustration his “Creativity Machine” algorithm produced. The “human authorship” component, which the USCO deemed completely necessary to establish a copyright, was missing. Thaler contended that this requirement was unlawful, but the USCO has steadfastly adhered to a century of copyright law. 

Ryan Abbott, Thaler’s attorney, intends to challenge the judgment. Requests by Richard Thaler for an AI named DABUS to be recognized as the creator of two goods were all denied by the US Patent and Trademark Office, the UK Intellectual Property Office, and the European Patent Office.

 

Thaler has filed appeals against those rulings, but case precedent has led courts to find that non-human expression is ineligible for copyright protection. The ruling cites earlier decisions from lower courts, including one from 1997 that determined a book of supposed divine revelations lacked the element of human arrangement and curation required for protection and one from 2018 that decided a monkey could not file a copyright infringement lawsuit.

AI generated image of railway tunnel
Stephen Thaler's AI creation, A Recent Entrance to Paradise, has been denied copyright protection by the US Copyright Office. Stephen Thaler

 

The requirement of human authorship for protection is less important in other nations. AI-produced ideas can be eligible for patent protection, according to a judge’s decision in Australia last year. South Africa also let Thaler patent one of his items the previous year, noting that “the invention was autonomously developed by an artificial intelligence.”

Artificial or Artistic? Concerns about ownership and copyright infringement in AI-generated art

To collaborate with other artists and teach machines to produce various types of visual works, artists create autonomous robots. Artificial intelligence has long generated art, but this year’s software innovations made it possible for even the least skilled artists to create complex, abstract, or lifelike works just by entering a few lines into a text box. 

An OpenAI model called CLIP (Contrastive Language-Image Pre-training), which serves as the primary link between text and images, is used to learn DALL-E 2. Data-trained AI can now independently produce photos and other works of art. The training data is an amalgamation of enormous datasets of tagged photos that have been categorized into a variety of categories, the majority of which are probably covered by copyright.

The resultant images have a figurative appearance, giving the impression that a real person or artist could have created the piece. Elon Musk, Greg Brockman, Ilya Sutskever, Sam Altman, Wojciech Zaremba, and Carlos Virella launched the artificial intelligence research facility known as OpenAI in 2015. In 2018, OpenAI unveiled DALL.E, a text-to-image creation model based on Transformers architecture that took inspiration from both the Wall-E robot and surrealist painter Salvador Dali. The software can create visuals in various styles, manipulate and rearrange the objects in those images, and precisely create fresh compositions without explicit guidance.

Additionally, it has shown that it can convey spatial and chronological information and solve Raven’s Matrices. The model’s ability to create well-known entities, such as trademarked logos and characters protected by copyright, as well as its ability to make realistic and context-aware edits, such as inserting, deleting, or retouching particular areas of an image from a natural language description, are its most crucial features.

Copyright, Contracts and Commercialization

The popularity of artificial intelligence-generated art is rising, and OpenAI has announced that it will be making DALL.E 2 and its image production platform available for purchase. However, AI-generated content presents several legal questions, including who is the rightful owner of these outputs and if anybody but a human may own the copyright. The Naruto Case clarifies the situation and provides an answer to this query.

British wildlife photographer David Slater unattended placed up a camera on the island of Sulawesi. Naruto, the monkey, pressed the button while posing for pictures and showcasing his photogenic side. Selfies and other images taken by Naruto were included in a book that Slater released. PETA (People for the Ethical Treatment of Animals) complained about Slater and the publisher on behalf of Naruto, claiming that he was entitled to own and profit from the Monkey Selfies’ copyright in the same way and to the same extent as any other author. If it is determined that a human being did not create the work, the Copyright Office declares that they will decline to register a claim. 

The phrase “original works of authorship” in Section 306 of The Copyright Act denotes that a human was involved in the creation. As a result, AI-generated art has no owner according to the laws as they stand. Artwork is protected by copyright as soon as it is created, giving the creator the only authority to determine how it will be used. The copyright issues that might be raised by photos produced by DALL.E 2 are the most crucial information in this work. 

These worries include copyright infringement charges on the final image based on works of art with copyrights fed into the AI during machine learning. Although the AI’s owner may be held accountable for infringement, it is likely that neither the AI nor the AI corporation owns any rights to the image because copyright law does not cover AI-generated artwork. The conditions of use specify that input and output are collectively referred to as “Content” and that OpenAI owns all rights, titles, and interests in and to output. Creators can submit an image to have it transformed by AI or input a prompt to make an image. The Services may generate the same or comparable output for OpenAI or a third party, and output may not be unique across users.

If all conditions are met, OpenAI grants the creator the right, title, and interest in and to the final image. The owner or assignor of a copyright may transfer ownership of that right to a third party. Regardless of whether or not the photographs have copyright, the author now owns them. The sharing and publication policy states that authors who wish to publish their first-party written content that was partially created using the OpenAI API may do so as long as the following requirements are met: the use of AI in the creation of the content must be explicitly acknowledged in a way that no reader could miss it, and a human must bear final responsibility for the content being published.

AI generated image of dog with purse (no copyright infringement)
AI generated image of dog with sunglasses
Painting of dog with sunglasses

Credit: DALL·E 2; text prompt: Oil painting of Gucci

 

OpenAI is taking precautions to reduce copyright infringement when utilizing DALL.E 2-generated artwork. These precautions include rejecting image uploads with recognizable faces, rejecting generation prompts that try to mimic the likeness of public figures, enhancing filters to prevent users from posting harmful or prohibited content, removing relevant data from the software’s training itself, implementing a new technique to improve 12x the generation of diverse images of people, and using both automated and human monitors to watch over the platform and prevent abuse. 

 

A graphic created from a word prompt that includes the name brand “Gucci” illustrates this. When specific terms are entered into the software, alternate images that avoid any recognizable company name, logo, or trademark may have been included but still produce intriguing or humorous artwork. Users who utilize DALL.E are granted complete commercial rights to the photos they produce, including the ability to sell, reproduce, and merchandise them. The terms of service further inform users that OpenAI has the right to modify these Terms at any moment and to pause or stop their use of the Services. Due to the lack of authorship or paternity on the part of the AI itself and the possibility of violating datasets that the AI has learned, this could raise legal issues.

Implications for copyright law

Artificial intelligence could have significant effects on copyright law while producing works. Because the programme was only a tool to aid in the creative process, copyright ownership in computer-generated works was traditionally uncontested. With the most recent developments in artificial intelligence, however, the computer programme is no longer just a tool; instead, it actively makes many of the decisions involved in the creative process without human input.

 

While they may be regarded as copyright-free because they were not produced by a human author, allowing for their unrestricted use and reuse by anybody could have significant business ramifications. As a result, developers may become less willing to invest in automated systems as they question if works produced through machine learning are worthy of copyright protection. With the reduction in labour expenses, using artificial intelligence to undertake laborious tasks might still be acceptable.

Legal options

According to copyright law, computer-generated works are not entitled to copyright protection or the authorship of such works is attributed to the program’s developer. According to a statement from the US Copyright Office, “given that the work was generated by a human being,” it will “record an original work of authorship.” The Court of Justice of the European Union (CJEU) asserts that copyright is only applicable to original works in Europe, and that a work must be the author’s intellectual creation in order to be called original.


In the UK, the author of a computer-generated literary, dramatic, musical, or artistic work is deemed to be the one who makes the arrangements required for the work’s development, according to section 9(3) of the Copyright, Designs, and Patents Act (CDPA). A computer-generated work is produced by a computer under conditions where there is no human author, according to Section 178 of the CDPA. By recognizing the labor that goes into designing a programme capable of generating works, even if the computer does the creative spark, this clause aims to offer an exemption to all human authorship criteria.


The line between artwork created by a person and that created by a computer is likely to grow even hazier as artificial intelligence is used by artists more frequently and as machines become better at producing creative works. The UK’s policy appears to be the most effective for ensuring businesses continue to invest in the technology. The next contentious topic is whether or not computers should be accorded the same status and rights as people, but that is another discussion.

This article attempted to forecast how existing copyright laws will apply to works created by AI. It presented two hypotheses to illustrate who might be the rightful owner of any AI-created works of art: the principal can retain copyright ownership by treating the AI as a principal’s agent and applying agency-principal rules, and the copyright should belong to the user rather than the AI developer since the end-user has ultimate control over what the AI produces. The article also included analysis of tort law, including right to publicity, to address torts perpetrated by AI. 

 

The future of legal protection for these works is still uncertain, but creators will likely be able to use DALL.E 2 and produce images that can be used commercially without worrying about facing legal action. The programme developed by Stable Diffusion supports a variety of user inputs and has a thorough explanation of how it works.
The inventor of DALL.E 2 has access to the source code and the ability to enter text prompts and combine ideas to produce the work. Machine learning algorithms are learning from the data provided by programmers and making their judgements along the way to decide how the new work will appear. The patent community is debating whether an AI can be identified as an “inventor” on a patent application in other areas of intellectual property.

Share This Post

Subscribe To Our Newsletter

Get updates and learn from the best

More To Explore

Doctor using AI for better therapy decision making

How can AI help with therapy in healthcare?

How can AI help with therapy in healthcare? The application of embodied artificial intelligence (AI) in psychiatry, psychology, and other mental health services like psychotherapy

How does Maya do it?

Stay ahead of the market while it’s happening

Meet Maya, a private generative AI for internal and external data insights