AI Copyright Roundup: Week in Review
Data Laundering. Credit Alec Foster, made with Midjourney and Photoshop

AI Copyright Roundup: Week in Review

It's been a huge week for AI and intellectual property law. In this update, I explore the latest developments concerning copyrighted works in AI datasets, a controversial practice known as AI Data Laundering, AI-generated artwork copyright, and Creative Commons (CC) licenses.

1. Plagiarism in AI Training Data

The recent exposure of pirated books utilized to train generative AI models, as detailed by Alex Reisner 's expansive piece for The Atlantic, sheds light on an alarming and ethically questionable practice. This revelation highlights the following key aspects:

  1. Unauthorized Usage: Thousands of works of renowned authors were found in the training data without proper licensing or attribution.
  2. Impact on Trust: Trust in AI systems is vital for their acceptance and integration into society. The use of pirated content in AI training can erode this trust, leading to commercial skepticism and resistance towards AI technologies.
  3. A Call to Action: The incident serves as a wake-up call for developers, regulators, and stakeholders in the AI community. Ensuring transparency, ethical compliance, and responsible use of data in AI training must be paramount.

2. AI Data Laundering

The practice of Data Laundering is emerging as a critical concern. It refers to masking the origin of unethically acquired data to make it appear legitimate.

  1. Case Study – Stability AI: Stability AI's use of data laundering involved scraping publicly available information and masking its origins. This allowed them to access vast amounts of data without respect for legal constraints. It reveals a gaping hole in our current regulatory framework.
  2. Academia and Nonprofits' Role: The Waxy.org investigation into AI data laundering goes further. It uncovers how academic and nonprofit researchers, often in collaboration with tech companies, can exploit data without proper attribution. This practice undermines the trust and integrity of AI research.
  3. Possible Solutions: Enforcing stringent regulatory measures, creating an oversight body, and implementing industry-wide standards could provide a framework for preventing this practice.

3. New Ruling on AI Artwork Copyright

In a ruling on Friday Aug 18, a federal judge upheld the guidance from the U.S. Copyright Office that a piece of art created by AI is not open to protection. I discussed this guidance from the USCO in my Aug 9 newsletter, "AI Copyright 101: A Comprehensive Guide".

A significant question has arisen: at what step between AI completely making a piece of art, and a human using AI for a very small piece in a much larger workflow is it considered "AI generated"? This presents a unique dilemma, such as if AI art can't be copyrighted, why can a painting made by a bucket dripping over a canvas be? Or if Photoshop's Content-Aware Fill feature on a photograph renders it public domain?

Some might argue that AI creations are the result of human coding and prompting, and should therefore be eligible for copyright. Others might view AI as a tool rather than a creator, leading to further complexity in legislation.

These pressing issues hint at a need for revised copyright legislation to integrate AI, providing clear guidance on how much human input is needed to be considered as a new piece of art, and not merely a recombination of existing art. Ultimately, I believe that once AI-generated art becomes convincing enough for media companies to utilize, new copyright laws will be enacted. Historical examples like the Copyright Term Extension Act of 1998 show that copyright guidance may be updated in line with business interests.

4. Creative Commons and Generative AI

Creative Commons (CC) licenses offer a flexible alternative to traditional copyright. Unlike full copyright, which reserves all rights, CC licenses allow creators to define specific permissions for how others may use their work. Ranging from very open licenses that only require attribution to more restrictive ones that forbid commercial use or derivatives, CC provides a spectrum of options that encourage legal sharing and collaboration.

The Intersection with Generative AI

Generative AI, with its ability to create new content from existing data, raises questions within the CC framework. Key challenges include:

  1. Attribution and Originality: Determining rights and attributions for AI-generated work that draws from human-created pieces.
  2. AI Training: Addressing legal and ethical dilemmas in using CC-licensed content for AI training.
  3. Responsibility and Transparency: Emphasizing transparent communication around origin and usage rights, aligning with Creative Commons' philosophy.

Generative AI's use of CC-licensed content brings a wealth of creative possibilities, but it also opens a Pandora's box of legal and ethical questions. Whose rights should prevail in this new creative landscape? How can we ensure fairness, creativity, and legal integrity in the age of generative AI? These are pressing questions that require a collaborative effort to address.

For more details, refer to the FAQ on AI and CC licenses.

Personal Reflections

Reflecting on the intersection of AI and human creativity, the challenges and potential are profound. The disconnect between artist and art in generative AI resonates with other forms of abstract art. While recognizing the need for clear guidelines and responsible practice, I am optimistic about AI's potential to redefine our creative landscape.

  • Historical Lessons: Historical precedents like the monkey selfie case provide valuable insight. It's not that AI-generated artwork cannot be copyrighted; it's a matter of recognizing the human guidance or ownership.
  • Ethical AI Training Model: The discovery of pirated books in AI training signals an imminent shift towards a model that rewards writers and artists, echoing a transition similar to the one from the Napster era in music to today's streaming model.
  • Caution and Hope: The dynamic intersection of AI-generated art and copyright juxtaposes caution with optimism. Though the legal entanglements are numerous, the horizon offers hope for copyright laws that will embrace AI's inventive potential while rewarding rightsholders.

As we navigate these complex issues, collaboration, thoughtful regulation, and an open mind will be key to unlocking AI's full creative potential.

Vas Bakopoulos

SVP | Brand Strategy, Data & Attribution, Marketing Insights | MMA, Possible, Digitas, Kantar, ARF, I-com | Instructor at NYU | Keynote speaker |

1 年

Thanks for sharing Alec. One of the analogies that I have heard for this topic is the "snake eating its tail", e.g. what happens when you train a model for academic purposes, and then as it happened with your case study someone repurposes it for commercial use cases etc. As the snake keeps eating its tail things get very murky about the underlying data and content IP. There are also interesting implications for patents and IP, a few of them blew my mind in this podcast https://www.audible.com/pd/Legal-consequences-of-generated-content-Podcast/B0CC6F63R4?source_code=ASSORAP0511160006&share_location=player_overflow

要查看或添加评论,请登录

Alec Foster的更多文章

社区洞察

其他会员也浏览了