AI ASSURANCE

AI System Self-Classifier.

Additional page sections

Walks through intended use, autonomy, affected persons and context to produce a preliminary AI system classification packet.

Version 1.5 Beta Protected engine AI classification decision tree artifact
PURPOSE

Decision supported.

Walks through intended use, autonomy, affected persons and context to produce a preliminary AI system classification packet.

Intended user

research, assurance and technical review teams

Output status

Preliminary outputHuman review requiredNot certification
USE CASES

Where this instrument fits.

  • Start AI governance intake
  • Identify high-risk indicators that need legal or policy review
  • Find missing information before formal classification
  • Create a preliminary classification note
INPUTS

Required input fields.

  • Intended use (required): Administrative support, Business decision support, Human-impacting decision support, Regulated or safety-relevant use
  • Domain (required): General enterprise, Employment or education, Financial services, Health or medical context, ...
  • Autonomy level (required): Assistive, Workflow automation, Tool-using or agentic
  • Human oversight (required): Meaningful review before action, Periodic review, Weak or unclear oversight
  • Data sensitivity (required): Public/synthetic, Internal, Personal, Special category or regulated
  • Supplier dependency (required): Internal, Mixed, External opaque model/provider

Data handling: this interface uses the L2ET protected same-origin instrument engine. Do not enter confidential, regulated, privileged, incident, medical or sensitive operational data.

METHOD

Decision Tree logic.

Decision-tree scoring combines intended use, domain, autonomy, oversight, data sensitivity and supplier dependency to flag indicators requiring further review.

Source families

EU AI Act source familyNIST AI RMFISO/IEC 42001internal risk taxonomy

Assumptions

  • Rules are intentionally conservative.
  • Jurisdiction-specific classification requires legal review.
  • User-entered context may omit important facts.
INTERACTIVE INSTRUMENT

AI classification decision tree artifact.

Use the controls below to generate a preliminary artifact. The output is intentionally bounded and requires human review.

OUTPUT ARTIFACT

AI classification decision tree artifact.

The generated artifact includes findings, assumptions, limitations, recommended next actions and exportable structured output.

Export options

Copy outputMarkdownJSON
EXAMPLE

Example input and output.

Example input

External GenAI system used for customer-facing recommendations with personal data and periodic oversight.

Example output

Flags high-risk indicators, missing evidence and human/legal review requirement.

LIMITATIONS

What this tool does not do.

  • Not a legal determination.
  • Not a regulator-grade classification.
  • Not a substitute for policy owner approval.

This instrument does not provide legal, medical, cryptographic, engineering, regulatory or compliance certification.

RELATED METHOD

Method and workflow links.

Read the family method note for assumptions, output artifacts, update policy and review boundaries.

Open methodology Open family

RELATED TOOLS

Suggested workflow sequence.

Assurance Radar

Measures eight AI assurance dimensions and produces a maturity radar with evidence gaps and next artifacts.

open instrument →

Evidence Reactor

Converts AI assurance claims into a Claim to Evidence to Control to Test to Owner review pipeline.

open instrument →

AI Governance Evidence Mapper

Maps an AI use case to the governance evidence required before approval, release or expansion.

open instrument →
CHANGELOG

Version history.

  • v1.5 - Research-grade instrument template, method notes, assumptions, limitations, example and export actions added.
  • Last updated: 2026-05-27.
  • Maturity state: Beta.