See it in action

Tell us who you are and what your team keeps asking. A person replies to find a time. You don't need to prepare anything.

We use these details only to reply to you.

Knowledge Vault and the EU AI Act.

The EU AI Act asks for records, human control and traceable AI output. This page maps each requirement to the part of the stack that handles it, and lists what stays with you.

Requirement by requirement

ArticleWhat it asksWhere in the stackHowCoverage
Art. 12Record-keepingHigh-risk AI systems log events automatically over their lifetime, so their behaviour can be traced.ReactorEvery change is an operation in an append-only log, with a timestamp, the signer's key and the app that made it. You can replay any past state.Built in
Art. 14Human oversightPeople can understand the system, step in and override it.Document models, SwitchboardAgents draft. Notes stay drafts until a named reviewer approves them, and authors can't approve their own. You can restrict approval to specific people.Built in
Art. 26Deployer dutiesDeployers keep logs for at least six months, give oversight to competent people and monitor use.Reactor, SwitchboardEach document keeps its log for its whole life. You grant access per vault (read, write, admin), and the Activity view shows who did what.Built in
Art. 11Technical documentationProviders keep technical documentation (Annex IV) and keep it current.Document modelsDocumentation lives as versioned documents. You can read any earlier version and export notes as Markdown or JSON.Supports
Art. 10Data governanceTraining and input data have a documented origin and quality checks.Provenance, sourcesEach note records its author and origin (import, manual, extracted or session) and links to the source it came from.Supports
Art. 13Transparency to deployersDeployers get clear information on what the system does and where it fails.Vault chatAnswers cite the notes they used. Instructions and known limits can live as reviewed notes.Supports
Art. 50Transparency of AI outputPeople know when content comes from AI.Reactor signaturesEach operation records the app and key that signed it, so you can tell an agent's changes from a person's.Supports
Art. 9Risk managementProviders run and document a risk process for the whole lifecycle.Tensions, health checksRisks and objections stay on record. Contradicting notes open a tension that stays open until a reviewer settles it.Supports
Art. 15Accuracy, robustness, cybersecuritySystems resist errors and attacks.Document models, SwitchboardDocument models reject malformed changes. Every write needs an authenticated, signed identity. You can self-host and pick a local model.Supports
Art. 4AI literacyStaff who use AI understand it.VaultYour procedures and decisions about AI use stay readable for every new joiner.Supports

What stays with you

  • Your partClassifying your AI systems by risk
  • Your partConformity assessment and CE marking (Art. 43, 48)
  • Your partRegistration in the EU database (Art. 49)
  • Your partFundamental rights impact assessment (Art. 27)
  • Your partTesting your models' accuracy
  • Your partDeciding who is accountable

Most duties for high-risk systems apply from 2 December 2027, after the 2026 Digital Omnibus.

This page explains how the product's features relate to the Act. It isn't legal advice. Whether you comply depends on your AI systems, your role and your processes.

Read the Act on EUR-Lex