EU AI Act readiness: classify your AI systems and build to the deadlines
We list your AI systems, suggest a provisional risk tier for each for your counsel to confirm, and build the logging, documentation and human oversight the EU AI Act asks for. Engineering work, done alongside your legal counsel.
- A free readiness review: a 30-minute call and a written follow-up
- High-risk obligations for Annex III systems from 2 December 2027
- Technical controls, not legal advice
Book a free readiness review
Key dates
The EU AI Act's dates after the 2026 Digital Omnibus
The EU AI Act is Regulation (EU) 2024/1689. The Digital Omnibus on AI, Regulation (EU) 2026/1744, in force since 27 July 2026, amended it and set the high-risk dates below. Obligations apply in stages, so the work for each AI system depends on its risk tier.
1 Aug 2024
The AI Act enters into force
Regulation (EU) 2024/1689 enters into force. Its obligations apply in stages over the following years.
What to do
Start an inventory of the AI systems you build, buy or use, so your counsel can confirm a risk tier for each one.
2 Feb 2025
Prohibitions and AI literacy
Prohibited AI practices are banned. Providers and deployers must take measures to support AI literacy among the people who run AI systems.
What to do
Check the inventory for prohibited practices with your counsel, and record how the people who run AI systems are trained.
2 Aug 2025
General-purpose AI models
Obligations for providers of general-purpose AI models apply.
What to do
Record which general-purpose models each of your systems uses, so your counsel can see where the model providers' obligations sit.
2 Aug 2026
Transparency under Article 50
Transparency obligations apply, for example to chatbots that talk to people and to systems that generate content.
What we build
Disclosure in chat and voice interfaces, and labels on generated content.
2 Dec 2027
High-risk obligations, Annex III
High-risk obligations apply to Annex III systems, such as those used for biometrics, critical infrastructure, education, employment and credit scoring.
What we build
Risk management, technical documentation, automatic logging, data governance and human oversight, in place before this date.
2 Aug 2028
High-risk obligations, Annex I products
High-risk obligations apply to AI in products covered by Annex I, such as machinery and medical devices.
What we build
The same technical controls, planned into the product roadmap now.
Not sure where to start?
Book the free readiness review. In 30 minutes we go through your AI systems and suggest a provisional risk tier for each, for your counsel to confirm, then follow up in writing with the technical work each tier needs.
Risk tiers
Four risk tiers, four kinds of work
The AI Act sorts AI systems by risk. The tier decides what you have to build, so the inventory comes first.
What the tier covers
- Prohibited: practices the Act bans outright
- High-risk: Annex III uses such as credit scoring, employment and education, and AI in Annex I products
- Limited risk: chatbots and AI-generated content, which carry transparency duties
- Minimal risk: every other AI system
Tier decides the work
What we build
- An inventory your counsel can check against the banned practices
- Risk management, technical documentation, automatic logging, data governance and human oversight
- Disclosure that people are dealing with AI, and labels on generated content
- Good engineering practice, and AI literacy for the people who run it
Dates and tiers per Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744, as of 16 November 2026. We build and document the technical controls. Legal advice stays with your counsel.
Deliverables
What the readiness work produces
What you hold at the end: the records the Act expects and the controls behind them, built into the system.
AI system inventory
Every AI system you build, buy or use, with its purpose, its users, the countries it serves and a provisional risk tier for your counsel to confirm.
Technical documentation
Architecture, data sources, testing and limits, written as the system is built and kept current with each release.
Automatic logging
Events recorded over the system's life. At iSEEit, every agent action is logged, with per-step traces of each model call.
See the iSEEit case studyHuman oversight
Who can stop or override the system, and how. At iSEEit, write-back to a customer's CRM requires human approval or a tenant-configured confidence threshold, recorded in an approval ledger.
Data governance
Training and input data checked for quality and bias. We sign a data processing agreement with clients whose data falls under GDPR.
GDPR-compliant developmentEvaluation and monitoring
Outputs scored against a test set before any prompt, model or tool change goes live, and performance tracked after release.
Dates per Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744, as of 16 November 2026. We build and document the technical controls. Legal advice stays with your counsel. Building AI now? See AI development in Europe.
FAQs
EU AI Act: common questions
Possibly, and your counsel should confirm it for each system. What we provide is the information that decision needs: an inventory of each AI system, what it does, who uses it and in which countries its output is used.
It depends on what it is used for, and your counsel makes that call. From 2 August 2026, Article 50 generally requires chatbots that talk to people to make clear they are dealing with AI. A chatbot used for an Annex III purpose, such as decisions on credit, employment or access to education, may be high-risk. In the readiness review we record what each chatbot does and who uses it, so your counsel has what they need.
Under the Digital Omnibus on AI, Regulation (EU) 2026/1744, in force since 27 July 2026, high-risk obligations for Annex III systems apply from 2 December 2027. For AI in Annex I products, such as machinery and medical devices, they apply from 2 August 2028. Article 50 transparency obligations apply from 2 August 2026.
A 30-minute call with an engineer to go through your AI systems and suggest a provisional risk tier for each, for your counsel to confirm. We follow up in writing with the key dates and the technical work each tier needs.
No. We build and document the technical controls: the inventory, logging, technical documentation, human oversight and monitoring. Whether a system is in scope, and what the law requires of you, stays with your legal counsel, and we work alongside them.
Many AI systems also process personal data, so we plan both together. We sign a data processing agreement with clients whose data falls under GDPR. Data stays in the EU, Swiss or UK region you choose, and one inventory records both the personal data each AI system uses and its provisional risk tier.