Charting Ethical Waters: The Dataset Providers Alliance’s Quest to Standardize AI Data Licensing

To go back to this article, navigate to My Profile, and then click on Saved stories to view them

A Fresh Initiative Seeks to Ethically License AI Data

The initial surge of significant AI technologies primarily utilized data that was deemed "publicly accessible" – essentially, all content that could be gathered from the web. Currently, entities providing data for training are starting to limit access and advocate for licensing contracts. As the search for more data sources grows more vigorous, innovative licensing ventures are springing up to ensure the steady supply of foundational material.

The Dataset Providers Alliance, an association established earlier this year, aims to bring standardization and equity to the AI sector. In pursuit of this goal, it has recently published a policy document detailing its viewpoints on key issues concerning AI. This consortium consists of seven companies that license AI, among them are Rightsify, a company specializing in the management of music copyrights, Pixta, a stock-photo platform from Japan, and Calliope Networks, a new venture in the field of generative AI copyright licensing. (The alliance anticipates welcoming at least five additional companies to its ranks come autumn.)

The DPA supports a system where consent must be actively given by authors and copyright owners before their data is utilized. This approach marks a notable shift from the practices of many leading AI firms. A number of these companies have established systems where the responsibility falls on the owners of the data to withdraw their content individually. Meanwhile, some companies do not allow for any form of withdrawal at all.

The Digital Privacy Alliance (DPA) advocates for its members to follow its policy of requiring consent, viewing this approach as significantly more principled. “It’s essential for artists and creators to agree to this,” states Alex Bestall, the Chief Executive Officer of Rightsify and Global Copyright Exchange, the music data licensing firm leading this initiative. Bestall regards the consent requirement as both a practical and ethical strategy, emphasizing, “Marketing datasets that are openly accessible is a quick path to legal trouble and loss of trust.”

Ed Newton-Rex, previously an AI leader and currently at the helm of the ethical AI organization Fairly Trained, criticizes opt-outs as being inherently unjust to originators. He points out that some creators might be unaware when such opt-out choices are made available. "It's especially encouraging to witness the DPA advocating for opt-ins," he remarks.

Shayne Longpre, who spearheads the Data Provenance Initiative, a group of volunteers that reviews AI datasets, appreciates the DPA's commitment to ethically obtaining data. However, he anticipates that the opt-in requirement might be difficult to implement due to the substantial amount of data needed for contemporary AI models. He suggests that this approach might lead to a scarcity of data or high costs, potentially making it feasible only for major tech corporations to afford the necessary data licensing.

In the document, the DPA expresses opposition to the idea of government-enforced licensing, advocating instead for a "free market" system where the creators of data and AI entities can negotiate directly. The alliance also offers more detailed recommendations. For instance, it proposes five different payment models to ensure that content creators and copyright owners are fairly compensated for their data. These models range from a subscription-based system, "usage-based licensing" where charges apply per usage, to "outcome-based" licensing where royalties are connected to profits. "This method could be applied to various mediums, including music, images, movies and TV shows, or books," Bestall mentions.

Bill Rosenblatt, a technology expert with a focus on copyright, believes that creating uniform pay structures could be beneficial. He points out that the Dataset Providers Alliance stands in an advantageous spot to establish these guidelines. According to Rosenblatt, it's crucial for AI firms to have motivations to embrace licensing agreements. Beyond the clear legal motivations like the fear of legal action or regulatory requirements for licenses, he emphasizes the significance of simplifying the licensing procedure. Rosenblatt argues that by making compensation mechanisms uniform, it can facilitate wider acceptance and use.

The DPA supports the utilization of artificial intelligence-generated synthetic data, predicting it will soon become the predominant form of training data. "There may be some resistance from copyright owners," Bestall comments, "However, it's bound to happen." The group is in favor of establishing clear licensing agreements for the initial data used in generating synthetic data and insists on clarity regarding its production process. Moreover, it urges for consistent "assessment" of these synthetic data models to address and reduce biases and ethical concerns.

Certainly, for the DPA to succeed, it's crucial to gain support from key figures in the industry, a task that's more challenging in practice than in theory. "Ethical data licensing standards are starting to take shape," notes Newton-Rex. "However, the adoption rate among AI firms is still lacking."

Nevertheless, the presence of the DPA signifies the conclusion of the unruly era of AI seems imminent. "The pace of change is rapid," notes Bestall.

Discover More…

Explore Political Insights: Subscribe to our newsletter and tune into our podcast.

Exploring the effects of providing individuals with unconditional cash assistance

Not all individuals experience weight loss with Ozempic.

The Pentagon is planning to allocate $141 billion towards a catastrophic device.

Occasion: Attend the Energy Tech Summit happening on October 10th in Berlin.

Additional Content from WIRED

Critiques and Manuals

© 2024 Condé Nast. All rights reserved. A share of revenue from products bought via our website may go to WIRED, as part of our Affiliate Agreements with retail partners. Reproducing, sharing, broadcasting, storing, or using the content from this site in any form is prohibited without explicit prior written consent from Condé Nast. Advertisement Options

Choose a global website