What is KlarvoEngine?

KlarvoEngine is Klarvo's regulatory intelligence engine. It analyses your AI system through a 3-pass classification pipeline and produces an article-cited compliance memo with obligations, deadlines, and recommended controls — in under 60 seconds.

What is the EU AI Act?

The EU AI Act (Regulation 2024/1689) is the world's first comprehensive AI regulation. It classifies AI systems by risk level and imposes obligations on providers and deployers of AI systems operating in the EU.

Any organisation that provides or deploys AI systems affecting people in the EU — regardless of where the organisation is based. This includes websites with AI chatbots, companies using AI analytics, and businesses integrating AI models into their products.

Prohibited AI practices have been enforceable since February 2025. GPAI model obligations since August 2025. Article 50 transparency and most remaining obligations apply from August 2, 2026.

What are the penalties?

Up to €35 million or 7% of global turnover for prohibited practices. Up to €15 million or 3% for transparency and other violations. For SMEs, the penalty is whichever amount is lower.

GPAI Model Provider Obligations

If your AI system uses a foundation model (GPT, Claude, Gemini, Llama, etc.), KlarvoEngine automatically checks for GPAI-related obligations under Chapter V (Articles 51-56).

What Is a GPAI Model?

A general-purpose AI model is trained with large amounts of data using self-supervision at scale, displays significant generality, and can perform a wide range of tasks. The EU Commission's guidelines add a technical threshold: models trained with more than 10²³ FLOPs that generate language, images, or video.

GPAI models: GPT-4, Claude 3/4, Gemini, Llama, Mistral, DALL-E, Stable Diffusion

NOT GPAI: Weather models, protein folding, game AI, transcription-only models

Two Tiers

Tier 1 — All GPAI Providers (Article 53):

Technical documentation (Annex XI)

Downstream provider information (Annex XII)

Training data summary publication

Cooperation with authorities

Tier 2 — Systemic Risk (Articles 51 + 55):

Additional obligations for models trained above 10²⁵ FLOPs:

Model evaluation with adversarial testing

Systemic risk assessment and mitigation

Serious incident reporting

Cybersecurity for model and infrastructure

Commission notification within 2 weeks

What Klarvo Users Need to Do

Most Klarvo users are DEPLOYERS, not GPAI providers. Your obligations:

Verify provider documentation — check your model provider supplied Annex XII docs

Check copyright policy — confirm provider has a published policy

Verify training data summary — check provider published this

Assess fine-tuning scope — if you fine-tuned, does your compute exceed 1/3 of original?

Document model version — record which version you use for traceability

When You Become a Provider

You become a GPAI provider if:

You fine-tune with compute exceeding 1/3 of original training

You develop a GPAI model from scratch

You place a GPAI model on the EU market for the first time

You do NOT become a provider if:

You use a model via API (you're a deployer)

You do prompt engineering or RAG

You build an AI system on top of a model (you're an AI system provider)

KlarvoEngine Detection

During classification, KlarvoEngine automatically:

Identifies if a GPAI model is mentioned in the description

Checks the model against known providers (OpenAI, Anthropic, Google, Meta, etc.)

Flags relevant GPAI obligations in your obligation list

Notes systemic risk status where known

Enforcement Timeline

GPAI obligations enforceable since August 2, 2025

GPAI fines applicable from August 2, 2026 (up to €15M or 3% turnover)

Pre-August 2025 models have until August 2, 2027 to comply

How KlarvoEngine Classification Works

3-Pass Pipeline Explained

Understanding Your Classification Memo