New nonprofit to certify AI models that only use copyrighted training data with permission
The launch of Fairly Trained marks a significant shift in the ethical landscape of artificial intelligence by establishing a certification for companies that respect copyright laws. Rather than allowing the industry to rely on the disputed "fair use" argument, this new nonprofit demands that AI developers obtain explicit permission or licenses before using copyrighted material for training. This initiative creates a clear market distinction between entities that operate with consent and those that exploit creative works without authorization. To qualify, companies must implement rigorous due diligence workflows and maintain transparent logs of their training data sources. This structure ensures accountability and provides a verifiable record of data provenance. By requiring these operational standards, the certification promotes a more sustainable ecosystem where intellectual property rights are honored, potentially reducing legal risks and fostering greater trust between technology developers and content creators. This development is crucial for the open data movement because it demonstrates that data sourcing can be both ethical and compliant with existing legal frameworks. It challenges the assumption that open access to data must come at the cost of creator rights, offering a model where transparency and permission coexist. For advocates of responsible AI, this certification provides a tangible benchmark for best practices, encouraging broader industry adoption of licensed and properly attributed datasets over unrestricted scraping.
Source: siliconangle.comPublished on 2024-01-19