AI Preferences (aipref) (datatracker.ietf.org)

The IETF has chartered a new working group called AI Preferences (aipref) in the Web and Internet Transport area to standardize how content owners express preferences about collection and processing for AI model development, deployment, and use. Chaired by Mark Nottingham and Suresh Krishnan (Area Director Mike Bishop), the WG will produce standard-track specifications that define a vocabulary for AI-related preferences and concrete ways to attach those preferences to content — either in metadata or via the protocols that deliver content. This effort aims to give websites, publishers, and other content providers interoperable primitives (think Robots-style signals) to declare whether and how their material may be used by crawlers, dataset builders, or models. Technically, the group will focus on vocabulary, transport/attachment mechanisms (e.g., Well‑Known URIs and HTTP response headers), and a standard method for reconciling multiple, potentially conflicting preference expressions. It will coordinate with IPTC/PLUS, WHATWG/W3C and other format bodies to enable adoption across HTML and media formats. Important boundaries: the charter excludes technical enforcement, authentication/authorization protocols, registries of preferences, and auditing. Drafts for the vocabulary and attachment specs are targeted for IESG submission by August 2026, providing a clear timeline for interoperable signals that could influence privacy, compliance, and data‑use practices in AI systems.