The dawn of Preference Signals: Going beyond licenses in the AI ecosystem
The popularity of proprietary generative AI systems trained on vast swaths of publicly available content has fueled concerns over exploitation, leading to a rapid decline of the AI data commons and litigation worldwide. Mechanisms like robots.txt are limited, and traditional copyright licensing has limited effects in the generative AI context. This presentation offers an overview of the collaborative, technical standardization efforts underway to establish machine-readable preference signals and shared ground rules for mutually beneficial data reuse. Standardization efforts are being driven by non-profit organizations like the IETF AI Preferences (AIPREF) working group and Creative Commons (CC). Industry-led solutions highlight the need for flexible commercial terms, including the Really Simple Licensing (RSL) standard, and Cloudflare’s AI Crawl Control, which facilitates pay-per-crawl and pay-per-inference licensing. The TDM·AI Protocol introduces ways to attach preferences to content identifiers. The presentation will also touch on legal contexts like the text and data mining (TDM) exceptions in European law.
The session addresses the critical intersection of Open Source, content rights, and the future of AI development. The "rapid decline of the AI data commons" directly threatens the core principle of open, collaborative development that defines the LF community. Open models require open data; without clear, standardized ground rules, data providers are increasingly resorting to restrictive measures, locking down content, and fueling litigation—all of which impede innovation. This session offers actionable intelligence for LF members by providing a comprehensive overview of the technical standards being developed. Attendees will gain the knowledge required to influence, integrate, and future-proof their AI strategy, ensuring that open technical standards are prioritized over proprietary, exploitative models.
Stefano Maffulli is a recognized authority on Open Source, with decades of experience as an enthusiast, contributor, start-up manager, and non-profit executive. He's currently the chief revenue office of Grist, the Open Source no-code platform used to build sovereign workflows. He served as Executive Director of the Open Source Initiative (OSI) - where he led the effort to define Open Source for Artificial Intelligence.