US AI pre-release review framework

US AI pre-release review framework
Event-driven SIPOC map auto generated by the ProcessHorizon web app

The US framework is not primarily regulating AI models as products but rather introducing a review process around the release of advanced AI models.

From a Points Architecture perspective, the voluntary pre-release review process can be decomposed into six governance dimensions, each represented as an event-driven SIPOC process.

Event-driven SIPOC is the process operating system, while the Points Architecture is the atomic governance layer that makes every transformation observable, controllable, auditable & optimizable.

1. AI Model Information Disclosure Review

Determine whether sufficient factual information exists to support a pre-release review. Datapoints Governance: Parameter count, Compute used, Benchmark scores, Safety test results, Security test results.

2. AI Risk Interpretation Review

Assess how different stakeholders interpret model risks. Viewpoints Governance: Developer viewpoint, Security viewpoint, National security viewpoint, Consumer protection viewpoint.

3. Release Authorization Review

Determine whether the model should proceed to release. Decision Points Governance of possible outcomes: Release approved, Release approved with conditions, Release delayed, Release rejected.

4. Model Release Lifecycle Monitoring

Govern the key events surrounding model deployment. Event Points Governance: Training completed, Review initiated, Approval granted, Model released, Safety incident detected, Model retired.

5. AI Safety & Security Controls Review

Verify that appropriate safeguards exist before release.                          Control Points Governance: Red teaming, Access restrictions, Model monitoring, Abuse prevention, Security testing.

6. Post-Release Observation & Learning

Capture evidence from operational use.                                       Observation Points Governance: Safety incidents, Security incidents, User complaints, Performance degradation, Emergent behaviors.

The US AI review framework is fundamentally a governance process. The model itself is not the governed object; rather, the chain of information, interpretation, decision, action, control & observation surrounding the model release is the governed system.

Using the following link you can access this sandbox SIPOC model in the ProcessHorizon web app and adapt it to your needs (easy customizing) and export or print the automagically created visual AllinOne SIPOC map as a PDF document or share it with your peers: https://app.processhorizon.com/enterprises/DSuFdpxvQP2K3CfcKrZDKWy4/frontend