The EU AI Act and Self-Hosted AI: What Changes in August 2026

Written by Jakub Rusinowski · Last updated 2026-07-12 · Hardware figures computed by our VRAM engine

This page is practitioner guidance on deployment mechanics, not legal advice — involve your counsel or DPO for decisions about your specific obligations.

The EU AI Act (Regulation (EU) 2024/1689) regulates AI by use case, not by hosting location — its main obligations for high-risk systems apply from August 2, 2026. Self-hosting does not exempt you from the Act; what it changes is role clarity and evidence: as the deployer (and sometimes provider) of an open-weight model you control the logs, documentation, and data flows the Act asks about, instead of depending on a vendor's paperwork. Here is the timeline, the role logic, and what an on-prem deployment does and does not solve.

The timeline that matters

The Act entered into force on August 1, 2024 and applies in stages (dates from the Regulation's own text):

DateWhat applies
February 2, 2025Prohibited practices (Art. 5) and AI-literacy duties
August 2, 2025Obligations for general-purpose AI (GPAI) *model providers*; governance structures
August 2, 2026The main event: obligations for high-risk AI systems (Annex III), transparency duties, most remaining provisions
August 2, 2027High-risk rules for AI embedded in regulated products (Annex I) and pre-existing GPAI models

Penalties scale to the violation: up to €35M or 7% of global annual turnover for prohibited practices, lower tiers for other breaches. The reason this page exists on an on-prem hub: the August 2026 date is when the compliance questions stop being theoretical for ordinary companies — and the survey data on the stats page shows procurement already reacting.

The two questions that determine your obligations

Question 1 — is your use case high-risk? The Act classifies by *what the system does*, not where it runs. Annex III lists the high-risk domains: employment decisions (CV screening, promotion), education scoring, credit and insurance eligibility, essential-services access, biometrics, critical infrastructure, law enforcement. An internal document-summarization assistant is not high-risk; the same model wired into hiring decisions is. Most business deployments described on this hub — internal assistants, RAG over documents, drafting — fall outside Annex III and face only transparency-level duties (e.g., users must know they're interacting with AI). The classification exercise, done honestly per use case, is the first deliverable.

Question 2 — are you a deployer, or also a provider? The Act splits duties between the *provider* (who develops/places the system on the market) and the *deployer* (who uses it under their authority). Self-hosting an open-weight model for internal use makes you a deployer of your system. Two escalations to know about: substantially modifying a high-risk system, or placing your own AI system on the market under your name, can move you into provider territory with its heavier documentation and conformity duties. And the *GPAI model provider* obligations (training-data summaries, technical documentation) sit with the model's maker — Meta, Alibaba, Mistral — not with a company that downloads and runs the weights; the Act also contains carve-outs easing some GPAI duties for open-source-released models. Where exactly your deployment sits is a counsel question; the point here is that the roles are defined and self-hosting doesn't blur them.

What self-hosting changes under the Act — and what it doesn't

What it does not change: your obligations. A high-risk use case carries the same deployer duties — human oversight, input-data relevance, monitoring, incident reporting, record-keeping — whether the model answers from your rack or a vendor's API. Anyone selling "on-premise = AI Act compliant" is selling a category error; run them through the vendor checklist.

What it genuinely changes — evidence and dependency:

The honest summary: the Act is workload regulation, and self-hosting is evidence infrastructure. Companies choosing on-prem "because of the AI Act" are usually choosing it for the same reason they choose it for GDPR — not exemption, but a shorter, self-owned path to demonstrating the things the law asks them to demonstrate.

A sensible preparation sequence

For a company running (or planning) self-hosted AI, the preparation that counsel will not object to:

1. Inventory and classify every AI use case against Annex III — most will land outside it; the ones inside get the full treatment. 2. Stand up the logging now — the gateway-with-audit-log pattern from the deployment guide satisfies the record-keeping instinct of both this Act and GDPR, and it's a day of work. 3. Pin and document model versions — which model, which weights hash, which system prompts, for each use case. Trivial to do on owned infrastructure; this becomes your technical file's backbone if a use case is high-risk. 4. Add the transparency touches — users told they're talking to AI; AI-literacy training logged (already applicable since February 2025). 5. Put the high-risk use cases through counsel before August 2026 — with the inventory, logs, and version documentation above, that conversation is short.

Frequently asked questions

Does running AI on-premise exempt a company from the EU AI Act?
No. The Act (Regulation (EU) 2024/1689) regulates by use case — a high-risk application carries the same deployer obligations regardless of where the model runs. What self-hosting changes is evidence and dependency: you control the logs, model versions, and data paths your documentation describes, instead of relying on a vendor's. Treat any "on-premise = compliant" claim as a red flag. This is practitioner framing, not legal advice.
When does the EU AI Act actually apply?
In stages: prohibited practices and AI-literacy duties since February 2, 2025; general-purpose AI model-provider obligations since August 2, 2025; the main high-risk system obligations from August 2, 2026; and rules for AI in regulated products plus pre-existing GPAI models by August 2, 2027. The August 2026 date is the one driving current enterprise preparation.
Is an internal LLM assistant "high-risk" under the AI Act?
Generally no — Annex III defines high-risk by domain: employment decisions, education scoring, credit/insurance eligibility, biometrics, essential services, law enforcement. A document-summarization or knowledge assistant falls outside those, facing transparency-level duties instead. The same model becomes high-risk the moment it's wired into, say, CV screening — classification follows the use case, so inventory per use case, not per model.
If we self-host Llama or Qwen, do we take on the Act's GPAI provider duties?
The GPAI model-provider obligations (technical documentation, training-data summaries) sit with the model's developer — Meta, Alibaba, Mistral — not with a company that downloads and deploys the weights, and the Act eases certain duties for open-source-released models. You remain the deployer of your system, with deployer duties scaled to its risk class. Substantially modifying a high-risk system or marketing your own AI product can escalate your role — that boundary is a question for counsel.
What are the EU AI Act penalties?
Tiered by violation type, per the Act's text: up to €35 million or 7% of global annual turnover for prohibited practices, with lower maximums for other non-compliance (up to €15M/3%) and for supplying misleading information (up to €7.5M/1%). SMEs face the lower of the fixed or percentage amounts. The practical takeaway isn't the ceiling — it's that the record-keeping that avoids trouble is cheap to build now.

Keep going

Rolling this out in your organization?

Jakub Rusinowski, the founder of LLM Configurator, runs corporate workshops and lectures on deploying local LLMs — hardware sizing, model selection, compliance-friendly architectures, and hands-on setup for your team. Direct, vendor-neutral, practitioner-level.

Ask about a workshop →