Your AI governance platform

    Klaveta protectsyour clientsfrom AI risks.

    Train your teams, map your AI usage, control your risks — and document your AI Act compliance journey.

    Test your AI risk level
    No credit card
    Backed by
    Village by CA RéunionMEDEF RéunionLa French Tech RéunionPépinière FNTC
    The problem

    Your employees use AI. You can't see what they put into it.

    Most SMEs are AI deployers under the EU AI Act: they use AI tools they didn't build.

    Your teams use ChatGPT, Claude, Copilot.
    Often with your data, or your clients' data.

    77 %

    of employees paste company data into generative AI tools — 82% of them via personal accounts.

    LayerX · 2025

    15 %

    of organizations have a written AI policy. Yet seven in ten say their staff are using AI.

    ISACA · 2024

    In most organizations, AI is used with no written framework.
    Article 4 requires your teams to master the AI they use. Which means knowing who uses it, for what, and on which data.

    • A client asks whether their file went through an AI.You don't know.
    • A manager has ChatGPT screen résumés.You find out too late.
    • An audit demands proof that your teams are trained.You have nothing.
    Concretely

    The cockpit for your AI governance.

    A single platform for your entire AI Act journey.

    01 / 04

    Train

    Each employee follows the path for their role. Short modules, a certificate at the end.

    • One path per role, 106 lessons
    • AI literacy certificate at the end
    • Automatic reminders for latecomers
    Test your AI risk level
    No credit card
    The scope

    It's not the tool that matters.
    It's what you do with it.

    You declare the use, Klaveta assesses the level and tells you what to document.

    ChatGPTone tool, three uses
    Meeting notes

    Internal meetings and summaries, no decision.

    Minimal risk

    Register entry, nothing more.

    Customer relations

    Replies to customers on the public website.

    Article 50

    The person must know they are talking to an AI.

    Recruitment

    Shortlisting applications.

    Annex III

    Enhanced documentation and human oversight, from 2 December 2027.

    Three uses, three levels, one vendor.
    It's the use that decides what you must document.

    Already covered
    ChatGPTClaudeCopilotGeminiMistralCanva IAPerplexityNotion IAExcel CopilotFirefly+your in-house tools

    Not sure which AI tools are already in use at your company? The check tells you in eleven questions.

    Test your AI risk level
    No credit card
    EU AI Act

    Part of it is already mandatory.

    What applies to you today, and what will apply next.

    Applicable today
    2 février 2025
    Article 4

    Appropriate AI literacy measures for any organization that uses AI.

    In force
    2 août 2026
    Article 50

    Transparency: users must know they are interacting with an AI.

    In force
    2 décembre 2027
    Annex III

    High-risk systems: recruitment, credit, education, essential services.

    Upcoming
    August 2, 2028
    Annex I

    AI embedded as a safety component of a regulated product.

    Upcoming

    Source: Regulation (EU) 2026/1744, OJEU of 24 July 2026.

    Your role

    Does this apply to me?

    The AI Act distinguishes two roles, and your obligations depend on yours. The same regulation does not ask the same of you depending on whether you use AI or build it.

    Deployer

    Most SMEs are here

    You use AI tools (ChatGPT, Copilot, a chatbot on your website…) you didn't build.

    Already in force
    • Build your teams' AI literacy

      Take measures to foster the development of AI literacy among the people who use AI on your behalf. A best-efforts obligation: the regulation does not require you to guarantee a given level.

      Art. 4 · since 2 February 2025
    • Use no prohibited AI practice

      Manipulation, social scoring, exploiting vulnerabilities: these practices are not brought into compliance — they stop.

      Art. 5 · since 2 February 2025
    • Tell people when it's AI

      A deepfake, or AI-generated text published to inform the public, must be disclosed as such — where it is read, not in your legal notice.

      Art. 50(4) · since 2 August 2026
    Upcoming (3)
    • New prohibited practices

      New prohibitions join Article 5, including non-consensual intimate imagery.

      Art. 5 — 2 December 2026
    • Manage your high-risk AI uses

      If you use AI to recruit, grade, grant credit or access essential services: follow the system's instructions for use, monitor how it operates, keep effective human oversight and inform workers' representatives before putting it into service at work.

      Annex III · Art. 26 — 2 December 2027
    • AI embedded in a regulated product

      AI used as a safety component of a regulated product (machinery, medical devices…).

      Annex I — 2 August 2028

    Provider

    You develop or market an AI system, under your own name or trademark.

    Already in force

    The same foundations as deployers — AI literacy (Art. 4) and prohibited practices (Art. 5) — plus:

    • Design systems that disclose they are AI

      A person interacting with your system must know they are talking to an AI — the system itself must inform them, by design.

      Art. 50(1) · since 2 August 2026
    • Mark the content your AI generates

      A machine-readable marking signals that the content is artificial. A visible disclosure is only required for deepfakes and content informing the public.

      Art. 50(2) · since 2 August 2026
    • Document your general-purpose model

      Only if you develop a general-purpose AI model (the engine, not the application): technical documentation, information for integrators, a copyright policy.

      Chap. V (GPAI) · since 2 August 2025
    Upcoming (3)
    • New prohibited practices

      New prohibitions join Article 5, including non-consensual intimate imagery.

      Art. 5 — 2 December 2026
    • Meet your high-risk provider obligations

      If your system is used to recruit, grade, grant credit…: risk management, technical documentation, human oversight built in by design, registration and CE marking.

      Annex III · Art. 16 et seq. — 2 December 2027
    • AI embedded in a regulated product

      AI used as a safety component of a regulated product (machinery, medical devices…).

      Annex I — 2 August 2028

    Both? That's the most common case: you use AI internally and ship it to your clients. Your obligations add up.

    Test your AI risk level
    No credit card
    What you put in place

    Compliance sets the rules.
    Ethics guides the decisions.

    Three pathways, 106 lessons in total. Everyone follows the one for their profile. The training that costs €10,000 in a classroom, included in your plan.

    Pathway 01

    Foundation

    The shared AI-literacy core, for anyone who uses AI at work.

    29 modules · 84 lessons
    Mandatory — AI literacy attestation
    Pathway 02

    Lead

    Compliance steering: risk mapping, registers and traceability, monitoring in production.

    9 modules · 20 lessons
    AI lead
    Pathway 03

    Option

    The add-on for regulated sectors: fundamental-rights impact assessment.

    1 module · 2 lessons
    Regulated sector

    Your obligations depend on what you do with AI. Let us see which ones apply to you.

    Test your AI risk level
    No credit card
    What you get

    Make your progress count.

    Five criteria covered, and the « Use AI • Stay Human » (« Use AI, stay human ») label activates — backed by a public page you can show to a client, an insurer or a contracting party.

    USE AI – STAY HUMANUTILISEZ L'IA – RESTEZ HUMAIN
    Criteria met5 / 5
    1. AI systems inventory kept up to date
    2. Literacy pathway completed and attested
    3. Usage rules published and read
    4. Maturity audit validated
    5. User information in place
    Backed by a public page

    Private label issued by Klaveta based on the measures you declare. It evidences your compliance process to your clients, partners and teams. At this stage, no vendor can issue an AI Act certification: the European regulation provides for none where a company uses AI. Klaveta structures, equips and documents your compliance process. Designed and hosted in France.

    What you publish

    A charter built on your actual AI use.

    Generated from your declared tools and their risk level. Versioned, read by the team, reviewed every quarter.

    1. Version 112/02 · 9 tools
    2. Version 204/04 · +2 tools
    3. Version 312/05 · in force
    Quarterly review scheduledExample
    AI usage charter
    Léa Conseil firm · version 3 · 14 tools declared
    In force
    1Approved and prohibited tools, by role
    2Data that never leaves the company
    3Mandatory human review before any decision
    4Informing the people concerned
    5Reporting an incident or a doubtful output
    6Review and update whenever a new tool is declared
    14 / 14 time-stamped reading confirmations
    LMTDKBRS+10
    Created in Réunion

    Accelerated at Village by CA. Selected by the Crédit Agricole ecosystem.

    Data hosted in Europe

    GDPR applies in full. Your data stays within the European Union.

    Framework kept current

    The regulation evolves, the platform keeps up. You have nothing to monitor.

    No installation

    A browser is all you need. Nothing to deploy, nothing to maintain.

    Solo
    You, with no employees
    99 € excl. VAT/year
    that is 8.25 € excl. VAT per month, paid in one go
    • A public trust page, kept current, that your clients can check
    • Your AI usage rules, written from the tools you actually use
    • Your inventory of AI uses
    • Your record of processing activities
    • The training path and your certificate of completion
    • Your PDF file, to send on request
    • Regulatory watch: what changes for you when the law moves
    • Your quarterly check-in and your annual review
    Join the waiting list
    Coming soon. We will let you know when it opens.
    Team
    You and your team, up to 49 employees
    5 employees1 to 9 employees
    90 € excl. VAT/month
    no commitment
    Every feature, whatever your band. Only the headcount changes.
    Everything in Solo, plus:
    • Everyone declares their own uses, straight from their invitation
    • Consolidation and reminders, without you chasing anyone
    • The staff information notice and the works council notice
    • Charter and training follow-up, person by person
    • An internal AI lead or an outside adviser, at no extra cost
    No commitment monthly. Two months free on annual.
    Enterprise
    From 50 employees
    On quote
    Let us talk first.
    • Everything in Team
    • Support from your IT department or your adviser on the steps that need it
    • A scoping session before you start
    Book a meeting
    Thirty minutes, no commitment.
    Prices excl. VAT — VAT added based on your location (8.5% in Réunion, 20% in mainland France, reverse charge in the EU).
    Public bodies and local authorities — on quote. Contact us →
    Launch offer

    Founding — 50% off your first year

    Reserved for the first 50 organisations, on a prepaid annual commitment.

    Video onboarding includedPriority supportDirect access to the founder
    Book a demo
    Thirty minutes, no commitment.

    Frequently asked questions.

    Resources

    Understanding the AI Act

    Pilotage & shadow AI

    ChatGPT, Claude, Copilot, Mistral : quel assistant IA pour une PME ?

    Le meilleur assistant n'existe pas en général. Il existe pour votre environnement, vos données et votre exigence d'hébergement. Cinq critères pour trancher.

    25 septembre 2026 · 9 min
    Pilotage & shadow AI

    ChatGPT et confidentialité : ce que deviennent vos données, offre par offre

    Sur un compte personnel, les conversations servent à l'entraînement sauf refus. Sur Business, non, mais l'administrateur peut les lire. Le détail.

    24 septembre 2026 · 8 min
    Pilotage & shadow AI

    Microsoft Copilot en entreprise : ce qu'il voit de vos données

    Copilot ne crée aucun droit d'accès : il rend visibles ceux qui existent déjà. C'est pour cela que le chantier commence dans SharePoint, pas dans Copilot.

    23 septembre 2026 · 8 min

    Your teams already use AI.
    It's yours to decide how.

    The assessment takes a few minutes and gives you your real obligations. No account, no credit card.

    Test your AI risk level
    No credit card