Licensing of your creative content, whether it’s a book, a painting, or an audio-visual work, for use by an AI system (for training ingestion and for AI output use) is the future – the near future. As a content owner of a creative work, your concerns can be summarized in 4 categories, i.e., Consent, Control, Credit, and Compensation.
1. CONSENT (Licensing). Right now, there are several companies that are licensing creative works from the owners for further licensing to AI Developers and systems for AI uses. These licensing companies get the rights from the content owners to license the creative works for AI uses of the creative works to AI developers. These companies include Copyright Clearance Center, Calliope Networks (www.calliopenetworks.ai), and Created by Humans (www.createdbyhumans.com). The licensing concerns you have as a creator of a creative work include:
- Opt-out. Can the owner/creator out-opt of AI Licensing Regimes?
2. Training Rights Licensing.
3. Output Rights Licensing.
-Outputs to be used for Internal Uses only. Such as internal market analysis, summarizing or synthesizing journals or databases, trend analysis, product discovery, R&D. No use of the creative work for outside (public facing) uses.
-Outputs to be used for External Uses. Includes all public facing uses for companies that want to commercialize the outputs for their clients.
4. Retrospective Licensing. Does the license cover previous unauthorized uses? This may include a waiver of prior unauthorized uses, including unauthorized ingestion into an AI system or removal of Technical Protection Measures (TPM) as to ingestion or scraping.
Many US and state laws have been proposed to address the issue of identifying the data and the works that are used to train their AI systems by ingestion and to identify and compensate the rights holders for use of the creative works in the outputs.
California is the first state to pass a training data law – AB 2013 – Training Data Transparency. California enacted into law a generative artificial intelligence (“AI”) law that requires developers such as CHATGPT and Gemini to post information on their websites regarding the data used to train their AI systems. Developers must comply with its provisions by January 1, 2026. The Developers must disclose the sources or owners of the datasets; a description of how the datasets further the intended purpose of the AI system or service; the number of data points included in the datasets, which may be in general ranges, and with estimated figures for dynamic datasets; a description of the types of data points within the datasets (e.g., types of labels used or general characteristics); whether the datasets include any data protected by copyright, trademark, or patent or whether the datasets are entirely in the public domain; whether the developer purchased or licensed the datasets; whether the datasets include “personal information” or “aggregate consumer information” as those terms are defined under the California Consumer Privacy Act; whether the developer cleaned, processed, or modified the datasets and the intended purpose of those efforts in relation to the AI system or service; the time period during which the data in the datasets was collected, including a notice if the data collection is ongoing; the dates the datasets were first used during the development of the AI system or service; and whether the generative AI system or service used or continuously uses synthetic data generation in its development.
2. CONTROL. This concern includes licensing but also focuses on INFRINGEMENT PROTECTION MONITORING & AUDIT rights ( to determine if the licensee is staying within the limits of the license, especially if the outputs cannot be publicly shared if that is a restriction in the license). This also includes the Audit Trail to follow how the licensed works have been used and the “provenance” of the works, which is discussed in depth under “Credit.
3. CREDIT (attribution). This focuses on digital trails that establish the human or non-human authorship of content; cryptographic metadata, watermarking, and fingerprinting of the original media content/object, digital watermarking, and online content identifiers to establish robust attribution trails.
This means that there needs to be a standardized system that can be applied to the many varieties of AI technologies and platforms. These systems are intended to create uniform “content credentials” for all types of content (additional metadata) that trace the provenance (original ownership and subsequent modifications to the content and the source) and include a digital signature in a tamper-evident package, i.e., the “content credential.” So, if a user clicks on the “content credential,” a user can see if it is AI created or partially AI created/generated and the attestation of authenticity, along with other information.
Creation of a universal standard for a “content credential”.
- The Coalition for Content Provenance and Authenticity (C2PA) https://c2pa.org/, has created and is further developing the standard called “C2PA.” This standard creates and attaches “content credentials” to creative works. Currently, content credentials are automatically attached to images generated through the DALL-E3 model, through Microsoft Designer, and Microsoft Paint. LinkedIn, a part of Microsoft, has adopted the C2PA credential. These content credentials will be added to Google Search and Ads and YouTube. And, the U.S. Department of the Defense (DOD) has a content credential associated with all photos and DOD has a branch responsible for that. NASA will be using C2PA as a way to label their content as authentic.
- HAND Human & Digital https://handidentity.com/. It is a DOI- Digital Object Identifier registration agency and business intelligence platform providing global talent identify resolution and verification standard. It works with Sony AFI and “Respeecher, https://www.respeecher.com/ (voice synthesis around unique identification of instances of identity of talent, of legal and natural people, connected digital replicas and fictional characters).
4. COMPENSATION (pricing that allocates for use of creative works in the ingestion and outputs of AI systems). Conceptually, the value of a work needs to be assessed and assigned and then allocated for the actual use in an output. Pro rata attribution models are being developed by ProRata.ai. NVIDIA/Getty is doing it as well as BIRA, an Israel company. These companies are really collective rights licensing agencies (including the previously mentioned Copyright Clearance Center, Calliope Networks (www.calliopenetworks.ai), and Created by Humans (www.createdbyhumans.com)).
In the EU, a new law that becomes effective in 2026, requires all AI systems and model developers to identify the works that were used to train their models.
The REAL PROBLEM: VALUATION. Compensation for the use of works not only for AI training/ingestion, but as used in the output need to be assessed and assigned. This is the REAL PROBLEM that will face the industry as a whole and will necessarily involve the U.S. Copyright Office.
For instance, if your book is ingested into an AI system, and it is used by the AI system to produce an output, how will you be compensated? Ingestion and output are 2 separate functions. For the output, how much of your work was used? Qualitatively, was your work the real basis for the output? How are few pages valued? A few notes of a musical composition? A few brushstrokes or a portion of a painting? Valuation and payment of the underlying works used in the creation of the AI generated output will be the most difficult task ahead of creators of works and the AI community. Valuation will also invoke the application of copyright laws including “fair use”.
Since about the 1940’s, this valuation has already been done in the musical composition and sound recording space on a worldwide basis. It is complicated – and that is an understatement. It involves lots of complicated legislation, the establishment of revenue boards, and a lot of US Copyright Office oversight.
I am looking forward to see how this will play out for all types of content in the AI space. I will continue to update you as things develop.
