Provenance: What It Takes to Prove Creator Data Dignity (medium.com)

In a recent development, Vektor Memory has initiated a proof-of-concept project aimed at establishing a reliable system for compensating creators whose work contributes to AI training. The team has conducted a series of rigorous tests to stress-test the voting mechanisms within their proposed arbitration system, uncovering significant vulnerabilities that could jeopardize its integrity. Notably, they found that open voting protocols were susceptible to manipulation by coordinated attackers, emphasizing the need for improved governance structures. To counteract this, the team replaced open voting with a randomly sampled jury, which significantly reduced the likelihood of biased outcomes. This project is significant for the AI/ML community as it attempts to address the complex issue of data dignity—a concept that argues human-created data deserves fair attribution and compensation when used for training AI models. By leveraging research such as the C2PA specification for content provenance and the Generative Content ID framework, the project outlines a multilayered architecture that includes mechanisms for tracking contributions, managing disputes, and ensuring fair payment. Although the first layer is operational as an open-source tool for registering creator work, the broader system is still in design, highlighting the challenges that remain in establishing a robust and equitable creator compensation framework within the evolving AI landscape.