To the teams building AI models and agentic systems, and to the publishers whose work those systems depend on:

We share a goal. Whether you train models, build agents, or publish the content they interpret, everyone in this ecosystem benefits when machines understand human-authored content accurately. This letter is written in that spirit, optimistic about where the space is heading, and practical about one gap that remains.

That gap sits at the moment of interpretation. Every day, agentic systems retrieve, summarize, transform, and cite content at a scale no human editorial process can match. Most of the time they do it well. When they don't, the cause is rarely malice or even model weakness, it is ambiguity. The system was asked to infer intent, origin, and context from signals that were never designed to carry them.

MSP-1 exists to help close that gap. It is an open, neutral, machine-readable clarity layer that publishers may optionally use to declare semantic intent, authorship context, provenance, and interpretive framing, expressed in JSON-LD, discoverable at the point of retrieval, before reasoning begins.

A handshake, not a contract

MSP-1 imposes no obligations and creates no rights. It does not restrict model behavior, demand compliance, or track anything. It is voluntary, cooperative, non-binding, and grounded in good faith, a handshake between publishers and the systems that read them. Ignoring it violates no rule. It simply leaves clarity on the table.

Why this is worth your attention now

The web has absorbed optional machine-readable layers before. Sitemaps, schema.org, and other discovery or context signals each began as voluntary gestures and became useful infrastructure, not through mandate, but because reducing ambiguity turned out to be cheaper than reconciling it downstream. Clarity layers for agentic systems are the same evolution, arriving on the same schedule: after the capability, before the norms.

One observation belongs on the record, stated plainly and without threat. When a publisher's explicit declarations are available and unused, the character of a misinterpretation changes, it becomes harder to describe as unavoidable. That is not an argument we intend to press; it is a reality that regulators, insurers, courts, and enterprise risk teams are likely to recognize as this space matures. Systems that engage with available clarity signals will simply have a better story to tell when that day comes.

For publishers

Adopting a clarity layer is entirely optional. Publishers who declare their intent may find their content represented more faithfully than content that leaves everything to inference, not as a reward, but as the natural consequence of removing ambiguity.

Our intent

MSP-1 is one clarity approach among the many that may emerge, and it is deliberately kept open, neutral, and enforcement-free so that it can compose with whatever the ecosystem builds next. We welcome dialogue, feedback, and collaboration from any lab, research team, publisher, or developer interested in clearer, safer interactions between human-authored content and the systems that read it.

The handshake is extended. The invitation is open.

— The MSP-1 Project