AI Training Data: The Evolving Landscape and Its Impact

The Misinformation of "Disappearing Data"

In the wake of AI advancements, concerns arose that training materials were rapidly disappearing. However, this claim is baseless. While creators are now explicitly requesting consent prior to use, vast amounts of data continue to be created daily.

"Consent in Crisis: The Rapid Decline of the AI Data Commons" argues that the demand for permission poses a threat to AI research. However, this view contradicts the fundamental principle that owners have the right to control the use of their property.

Why the Change Now?

Until recently, publishers were unaware of AI's potential use of their materials. Their general rights reservations did not explicitly address AI. However, the EU's recent copyright law change has incentivized rightsholders to limit such use.

However, these restrictions do not prevent materials from being used in AI applications.

The Market for AI Training Data

While OpenAI's deals with major publishers may gain attention, countless smaller companies also enter into licensing agreements to access content.

Licenses are customizable to cater to different market segments, including small and medium enterprises (SMEs). Differentiated pricing has been a common practice in licensing for centuries.

Availability of Data for Research

The distinction between "research" and "commercial" AI use is often blurred. Even non-profit organizations may engage in commercial AI activities.

However, publishers have historically supported non-commercial research use of their materials. Access to research-oriented data remains abundant.

Limitations on Available Training Materials

While online content is finite, offline content provides ample opportunities for AI advancement. Publishers are willing to license their materials on equitable terms.

The limitations on training materials stem not from data scarcity but from the unwillingness of some tech companies to fairly compensate creators for access.

Conclusion

Publishers are justified in protecting their valuable content and exercise due caution when granting access to their materials. AI companies must recognize the importance of respecting copyright laws and fairly compensating creators for the use of their content.

As the NY Times noted, "the web will start shutting its doors" if tech companies continue to exploit data without providing value in return.