Choose one route before the full library.
Most visitors need one record, not twenty articles. Choose the closest situation, finish that first step, then open specialist guides only if needed.
I need the legal baseline
Start with roles, risk tiers, current-law timing, and the first control areas.
Read baseline →We already use AI systems
Create inventory fields that route systems to duties, owners, vendor checks, Article 50, impact review, and evidence records.
Build inventory →Audit or buyer questions are coming
See reviewer questions and map each one to a file, owner, gap, and next action.
Open audit questions →Move from reading to a record.
Use this order: baseline, inventory, evidence route, then specialist checks. This keeps the guide library from becoming another unread shelf.
Read the baseline
Check the role, risk path, current-law timing, and whether the AI system needs deeper review.
Read baseline →Create the system record
Record owner, purpose, vendor, data, risk path, Article 50 signals, and evidence status.
Open inventory fields →Map the evidence route
Route one system to oversight, vendor, disclosure, impact review, incident, training, and audit records.
Build evidence route →Open the blocked specialist guide
Use vendor, Article 50, impact review, sector, or incident guides only when the evidence route shows they matter.
Start vendor evidence →Use these before browsing every topic.
Start with high-risk classification. Use the Article 6(3) filter record only when the router points there, then move to evidence, vendor review, or transparency records.
High-Risk Classification Router
Use the 19 May 2026 draft Commission guidance as source context for Article 6, Annex I, Annex III, and Article 6(3) triage.
Run router →Article 6(3) Filter Decision Record
Build the downstream Article 6(3) record only when the classification router points to possible filter review.
Build record →Deployer Evidence Route Builder
Route one AI system to inventory, vendor files, oversight, logs, Article 50 notices, impact review, incidents, and audit readiness.
Open builder →Vendor Evidence Request Builder
Generate supplier questions, document requests, follow-ups, and red flags before enabling or expanding an AI feature.
Start vendor review →Article 50 Implementation Pack
Create disclosure records for chatbot, synthetic content, deepfake, emotion recognition, and other user-facing transparency triggers.
Open pack →Article 50 Evidence File Guide
Build a deployer record for provider marking, deployer labelling, deepfake disclosure, public-interest text review, reviewer approval, and retained evidence.
Open guide →Use the full library by task.
The full guide set stays available, but it is grouped by task so first-time visitors do not face a wall of options.
Frequently Asked Questions
Start with the Audit Questions Hub for evidence, the Compliance Guide for the baseline, or the AI Inventory guide if your organisation uses AI but has no register.
No. The guides are educational resources for structuring evidence, governance, and internal review. Formal legal conclusions need qualified counsel.
Start with the inventory. Then classify each AI system, collect vendor evidence, assign oversight, check Article 50 triggers, route impact review where needed, and keep a gap register.
Sector guides translate EU AI Act duties into contexts such as finance, insurance, healthcare, education, HR, and national enforcement.
No. Guides explain the logic. Tools run a specific check, create a working record, or prepare evidence for internal review.
Check regulatory references before major decisions, policy updates, procurement, or high-impact AI deployment.
Turn the guide into a record.
Use a tool when you need a quick check. Use a guide when you need the logic. Use evidence starters when someone asks for the record behind the decision.
Educational only. Not legal advice. Validate legal, privacy, employment, procurement, and sector decisions with qualified professionals.