FilingsGPT · Safety Frameworks

What disclosure
already looks like.

Article 44-B will require covered developers to publish a safety & security protocol describing how they manage the risk of critical harm. No such filing exists yet — the regime takes effect January 1, 2027. In the meantime, several frontier labs have voluntarily published their own safety frameworks: risk thresholds, capability evaluations, deployment gating, and governance. The six below are shown as a reference for what a 44-B disclosure looks like.

Illustrative reference entries — voluntarily-published lab frameworks, not Article 44-B filings. For cross-lab comparison see METR’s Frontier Safety Policy index and the Frontier Model Forum’s framework catalogue.

Published frontier-safety frameworks

7 frameworks

DeveloperFramework44-B obligations mappedSource
AnthropicResponsible Scaling Policy (RSP) v3.22026-04-29
Frontier AI frameworkCatastrophic-risk managementCapability / risk thresholdsModel evaluations & testingDeployment gatingGovernance & oversight
PDF ↗
OpenAIPreparedness Framework v22025
Frontier AI frameworkCatastrophic-risk managementCapability / risk thresholdsModel evaluations & testingDeployment gatingTransparency & disclosure
PDF ↗
Google DeepMindFrontier Safety Framework v3.12025
Frontier AI frameworkCatastrophic-risk managementCapability / risk thresholdsModel evaluations & testingDeployment gatingGovernance & oversight
PDF ↗
MetaFrontier AI Framework v22025
Frontier AI frameworkCatastrophic-risk managementCapability / risk thresholdsModel evaluations & testingDeployment gating
PDF ↗
MicrosoftFrontier Governance Framework2025
Frontier AI frameworkCatastrophic-risk managementGovernance & oversightTransparency & disclosure
PDF ↗
xAIFrontier AI Framework Dec 31, 20252025-12-31
Frontier AI frameworkCatastrophic-risk managementCapability / risk thresholdsModel evaluations & testingDeployment gating
PDF ↗
AmazonFrontier Model Safety Framework v2.72025
Frontier AI frameworkCatastrophic-risk managementCapability / risk thresholdsModel evaluations & testingDeployment gatingGovernance & oversight
PDF ↗

Sources: each developer’s own published framework, linked above where a stable public document exists. These voluntary frameworks are reproduced as a reference for the kind of disclosure Article 44-B contemplates; they are not official Article 44-B safety & security protocols, which do not yet exist (the regime takes effect January 1, 2027). Verify any version number, date, or specific figure against the live source.

FilingsGPT is an independent reference implementation of New York's Article 44-B (the RAISE Act). It is not affiliated with or operated by the NY Department of Financial Services or the State of New York, and is not an official filing system.