The SCL AI Requirements Framework establishes verifiable, pass/fail requirements for each AI specific failure mode not addressed by existing safety critical software standards. Requirements are defined by the failures they prevent, not the capabilities they enable. Algorithm agnostic. Domain applicable. Openly published under a citable DOI.
A system evaluated on data it effectively trained on, reporting skill it does not have outside its declared operating domain.
Outcomes that are systematically worse for some users because the training data encoded a historical bias.
Strong average performance that hides untested edge cases, boundary conditions, and injected failure modes.
A model accurate at deployment quietly drifting wrong as the world changes around it.
An AI stating fabricated facts, policies, or values with the same confidence as correct ones.
The model meeting an input unlike anything it was trained on and responding as if it were routine.
Inputs, training data, or the model itself manipulated by an adversary to force a chosen failure.
A safety-critical decision that cannot be inspected, questioned, or justified after the fact.
Operators who over-trust, under-trust, or lose the authority to override the AI when it matters.
Personal or protected data exposed, retained, or transferred without lawful basis or control.
One model’s error cascading through a chain of models into a system-level failure.
Network-specific failures: miscalibrated confidence, brittle features, and unverified deployment transforms.
A model that keeps learning after deployment teaching itself the wrong thing, unnoticed.
The framework applies all thirteen requirement areas at different depths depending on the classification of your AI system. Core requirements (AI-1 through AI-10) apply based on classification tier. Architecture and paradigm requirements (AI-11 through AI-13) apply conditionally based on system design. Classification is determined during Phase 1 of the assessment.
Assessment against this framework produces one of two outcomes, accompanied by a quantified risk score. Each determination is documented against a specific version of the framework, at a defined classification level, with every finding on record. There is no maturity index and no subjective rating.
The framework is openly published under a citable DOI. Every requirement, verification method, and evidence standard is available for review before any assessment begins. Verification is co-located: every requirement carries its verification, and every verification traces to a requirement.
If you believe a requirement is technically incorrect, insufficiently grounded, or missing coverage for a known AI failure mode, SCL welcomes that challenge. The standard improves through scrutiny.
Every determination follows the same four phase process.
See the process