Skip to main content
v2026.11,858 entries · CC-BY 4.0
Track N8

Transparency and review

Redaction, redaction reason, external review, evaluator access attestation, publication rights clause, AI-model review, evaluator independence and conflict of interest.

External review only means something if the evaluator is independent and a developer can't quietly redact an inconvenient finding. This track defines evaluator independence and conflict of interest, evaluator access attestation, publication rights clause, and redaction -- the transparency vocabulary behind third-party evaluation work by METR, the UK AI Safety Institute, the US NIST CAISI, and the AI Evaluator Forum, crosswalked against Anthropic, OpenAI, Google DeepMind, xAI, Meta, and California SB 53.

  • Evaluator Independence and Conflict of Interest
    Proposedrecord-type

    A proposed record of declared financial, organisational and personal ties between an evaluator and developer, and the independence test used to clear them.

  • AI-Model Review
    Proposedrecord-type

    A proposed record that an AI system performed an assurance task, capturing its identity, task, access, oversight, and the accountable human owner.

  • Publication Rights Clause
    Proposedproperty

    A proposed structured field recording contract or policy terms on what an evaluator may publish, who may redact it, and any developer approval rights.

  • Evaluator Access Attestation
    Proposedrecord-type

    A proposed joint record of what access an evaluator was granted or denied, for how long, and under what confidentiality and data-handling terms.

  • External Review
    Proposedrecord-type

    A proposed record of an assessment performed by a party outside the model developer, capturing that review's type, scope, and output.

  • Redaction reason
    Proposedcontrolled-value

    A proposed controlled value naming why information was redacted from a published document, covering permitted reasons and reasons a developer may not cite.

  • Redaction
    Proposedrecord-type

    A proposed record of a removal of information from a published document, noting where it occurred, how much, why, who decided, and that it happened at all.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →