How do I get an EU AI Act risk classification for an AI system with an agent?

Steps

1. Install the Strale Python SDK, straleio.
2. Create the client with your API key.
3. Describe the AI system's purpose in the description field.
4. Read the risk level, obligations and the articles that matched.

The solution

# pip install straleio
from straleio import Strale, DoRequest

strale = Strale(api_key="sk_live_YOUR_KEY")

result = strale.do(DoRequest(
    capability_slug="eu-ai-act-classify",
    inputs={
        "description": "AI system that scores job applicant CVs "
        "and recommends candidates for interview"
    },
    max_price_cents=100,
))

print(f"Risk level: {result.output['risk_level']}")
print(f"Prohibited: {result.output['prohibited']}")
print(f"Obligations: {result.output['obligations']}")
print(f"Applicable articles: {result.output['applicable_articles']}")

How it works

The EU AI Act sorts AI systems into risk categories: prohibited, high risk, limited risk and minimal risk. The category decides which obligations apply.

The tool is rule-based and does not use a language model. It matches your description against the Annex III categories and the criteria in Articles 5, 6 and 50, and returns risk_level, prohibited, obligations, applicable_articles and a classification_reason that says which criteria matched.

Matching is by keyword, so describe the system in standard terms, such as "CV scoring", "biometric identification" or "credit scoring", and not in vague language. classification_reason shows exactly what triggered the result.

The result is a starting point. Whether a system falls under the Act is a legal question, and the tool does not answer it.

Tools in this guide