A newly proposed bill, the Nurture Originals, Foster Art, and Keep Entertainment Safe (NO FAKES) Act, is facing scrutiny from the open-source community for provisions that critics warn could effectively ban local AI models. While the legislation aims to protect individuals’ digital likenesses from unauthorized AI replication, a specific clause regarding technical watermarking and fingerprinting poses a severe risk to privacy and open development.
The controversy stems from the requirement that AI models must be identifiable at inference time. For large, centralized corporations, embedding a static signature into a model is manageable. However, for open-source models—often modified, quantized, and run locally by users—this creates an impossible compliance burden. If a user strips a watermark or modifies a model to remove a fingerprint, they could face severe legal penalties under the Act.
This mirrors the complex legal battles seen in the software industry, where even determining the provenance of code can lead to litigation. Critics argue that by mandating traceability in a way that is incompatible with the decentralized nature of open source, the Act could force local AI projects offline, leaving the market dominated solely by large tech firms that can afford the compliance infrastructure.
Leave a Reply