Using AI in your company? The AI Act already applies to you.
We train your staff and set up with you the register that Art. 4 expects you to be able to show: who you trained, on what, with what result.
Sample data. The real register is filled in during the programme.
The real timeline: what already applies and what does not
In 2026 the EU postponed the high-risk rules to fixed dates. Staff training and transparency remain in force.
Choose the profile closest to yours
Summary for information only. Individual systems are classified case by case during the programme.
Five things you have been told about the AI Act
These are the claims most often heard from consultants and vendors. Press “What the law says” to compare.
Who did you train, on what, with what result? The register answers in a minute.
Today: certificates scattered across a shared folder. To answer an inspection you have to rebuild everything by hand.
Art. 4, as rewritten in 2026, asks providers and deployers to “take measures to support the development of AI literacy” and states that it “does not require providers or deployers to guarantee any specific level”.
What you need is evidence of the measures.
Four modules, with clear topics and timing
From kick-off to dossier in six weeks
Example for a 25-person company with 3 managers and one AI lead. Timing and formats adapt to your organisation.
E-learning in four sessions plus the test, spread over the weeks.
Half a day live, in a virtual classroom or on site.
Two three-hour sessions on your register.
Indicative proposal. Schedule and formats are agreed in the first call.
At the end of the programme the register is already filled in
During the course you work on NexusOS. Training, the tool inventory and the dossier stay in one place.
AI literacy programme
Two tracks, each with sessions, a study guide and a self-assessment test.
Fundamentals · Risks and the AI Act · Responsible AI · Governance
Fundamentals · Risks and the AI Act · Responsible AI · Governance at work
Training register
For each person: track, dates, score, result. This is your Art. 4 evidence.
| Person | Track | End | Score | Result |
|---|---|---|---|---|
| A. Ricci | Staff | 14 Oct 2026 | 86 | Passed |
| M. Serra | Managers | 16 Oct 2026 | 88 | Passed |
| G. Fabbri | Staff | 21 Oct 2026 | 58 | To retake |
AI tools used by staff
The first inventory every company needs, without impact assessments where they are not required.
| Tool | Use | Class | Training |
|---|---|---|---|
| Generative text assistant | Quotes, emails | Limited | Covered |
| AI features in design software | Modelling | Minimal | Covered |
| Image generator | Website renders | Limited | Partial |
Guided classification of a use
Gaps, by urgency
| Item | Gap | Severity |
|---|---|---|
| Image generator | Notice on published images | Blocking |
| Generative assistant | No owner assigned | Warning |
| Training | 3 people still to complete | Warning |
Internal audit dossier
Tools in use, training, notices, providers and open gaps in a single document, in PDF or Excel. It is what you show an inspector, a client or your Model 231 supervisory body.
Illustrative screens with sample data. The platform is a register for organisations that use AI: it is not a notified body or a legal advice service.
In Italy, corporate liability linked to AI is changing
Legislative Decree no. 160 of 9 September 2026, published in the Official Gazette no. 214 of 15 September 2026.
Two AI-related offences
The new Art. 25-vicies provides for 600 to 1,000 quotas (about €155,000 – €1.5 million) for omitted security measures on high-risk systems and 200 to 700 quotas (about €52,000 – €1 million) for the unlawful dissemination of AI-generated content, plus disqualification sanctions. The value of a quota depends on the size of the company.
Injured parties are better protected
The court can order disclosure of evidence on how the system works. Compliance, even when certified, does not in itself exclude liability.
Document the measures
If you have a Model 231, update it. Recorded training is a measure you can demonstrate.
Art. 18: “the causal link between the violation and the damage is presumed, unless proven otherwise”.
Summary for information only, not legal advice. Updating your Model 231 requires your lawyer.
Today no one can certify your compliance with the AI Act.
There are no harmonised standards cited in the EU Official Journal yet, and no notified bodies. We tell you this before selling you anything.
No certification
Our certificate records attendance and the test result.
ISO/IEC 42001
A good organisational basis, but not a presumption of conformity.
No legal advice
When it is needed we tell you, and a qualified lawyer provides it.
What we provide
Method, training, documented evidence and a register that stays up to date.
The questions we are asked most often
We only use design software. Does the AI Act apply to us?
If you use AI features in your tools or generative assistants, yes: you must train your staff and disclose the generated content you publish. High-risk obligations do not cover design work, except in specific cases.
What penalties do we face if we do not train our staff?
The Regulation does not attach a specific fine to Art. 4. What matters is something else: in the event of an incident or dispute, documented training shows the company's due diligence, and since 30 September 2026 this carries more weight in Italy.
Our software provider is compliant. Is that not enough?
The provider is responsible for its own obligations: design, documentation, marking of generated content. Training staff, informing people and disclosing the content you publish remain obligations of whoever uses the tool.
From 2 December 2026, do we have to mark generated content?
No. Technical marking is an obligation of the provider of the generative system. Your part is to disclose the realistic content you publish, such as images and videos resembling real people, places or objects.
How much time does it take?
The first module is delivered as e-learning, with a final test. The full programme depends on the number of people and the tools you use: we define it together in the first call.
Who is responsible if we misclassify a tool?
Compliance remains the company's responsibility, as with any law. We provide the method, the training and the tool, and we tell you when legal advice is needed.
Try it with a tool you already use
The same five steps as the NexusOS guided classification. Indicative result: not a conformity assessment.
The first step is a thirty-minute call
Introductory call
Which AI tools you use, who uses them, which obligations apply to you.
A proposal based on your numbers
People to train, modules, format and timing, calculated for your company.
Kick-off
Training, register set up, dossier exported at the end of the programme.
Innovative startup based in Rome: governance, quality and AI for companies, public administration, universities and healthcare. Lean AI · Human in the Loop method.
Sources: Reg. (EU) 2024/1689 as amended by Reg. (EU) 2026/1744; Italian Law no. 132 of 23 September 2025; Legislative Decree no. 160 of 9 September 2026. Status as of 29 September 2026.
