Actors, models and scope

Developer, model identifier, deployment surface, coverage date, safety framework version, coverage scope threshold, severity threshold, risk domain.

Every AI safety framework has to answer the same first questions before anything else: which developer, which model, and as of what date? This track defines the scope elements -- developer, model identifier, deployment surface, AI model card coverage, and risk domain -- that Anthropic's Responsible Scaling Policy, OpenAI's Preparedness Framework, Google DeepMind's Frontier Safety Framework, xAI's Frontier AI Framework, Meta's Advanced AI Scaling Framework, and California SB 53 each define in their own way, so a claim made under one framework can be located against the others.

• Risk domain
  Proposedcontrolled-value

  A proposed top-level category of catastrophic harm, drawn from a controlled ladder of values, under which threat models and thresholds are grouped.

• Severity threshold
  Proposedproperty

  A proposed structured value for the harm magnitude a framework treats as in scope, with a magnitude, a unit, and any stated exclusions.

• Coverage scope threshold
  Proposedrule

  A proposed if-then test for whether a framework, statute, or programme applies to a given developer or model, such as a compute or revenue figure.

• Safety framework version
  Proposedrecord-type

  A proposed versioned record of a governing safety framework's identifier, effective date, and change log, tying artefacts to the exact version in force.

• Coverage date
  Proposedproperty

  A proposed date marking when a report's risk statements hold true, kept distinct from the report's own publication date.

• Deployment surface
  Proposedproperty

  A proposed term for the product, API, or channel through which a model is made available, distinct from its deployment type.

• Model identifier
  Proposedproperty

  A proposed persistent identifier for a specific model version or checkpoint, with a rule for which changes keep it stable versus require a new one.

• Developer
  Proposedclass

  A proposed definition of the legal person responsible for developing a model or publishing a safety artefact, identified per jurisdiction.