Ventrova · EU AI Act Annex IV · Obligation already in force since 2 Aug 2026

The Annex IV documentation obligation for high-risk AI systems is already in force. If you're covered and don't have it, you may already be non-compliant.

Under Regulation (EU) 2024/1689 Article 113, the Article 11 / Annex IV technical documentation obligation for high-risk AI providers applied from 2 August 2026, that date has passed. This isn't a future deadline to prepare for, it's a documentation gap that exists right now if you haven't filled it. Sentinel Scan actually tests your system (15 attacks, OWASP LLM Top 10-mapped), then turns the results into Annex IV Section 2/3/5/6/9 evidence: what was tested, what was found, what's still open. No login, no dashboard, no multi-step questionnaire to fill out first.

What happens next: you send the email above, we reply same business day with a Stripe invoice link, you pay and send your scan JSON (or authorize us to run it), you get your evidence within 24 hours, so you can close the gap fast rather than stay exposed while you figure out a bigger program. One-time payment, no subscription, no account to create.

Not sure the obligation applies to your system yet? See the Annex IV explainer first. Want to run the underlying scan yourself? Get the free Sentinel Scan CLI, or see the full managed audit for control mapping and a questionnaire answer bank.

We are not a law firm. This isn't legal advice, and we're not asserting your specific system is or isn't in scope, only that the Article 11 obligation for covered high-risk providers is already active law, not a future date on a calendar.

What's included

  • ✓ Annex IV Section 2/3/5/6/9 evidence slice, generated from your Sentinel Scan CLI run
  • ✓ OWASP LLM Top 10-mapped findings table as the underlying risk-management evidence
  • ✓ Unsigned attestation block for your team to complete after remediation review
  • ✓ Raw scan JSON included as a machine-readable evidence appendix
  • ✓ Delivered within 24 hours of your scan

What's explicitly not included (the honest boundary)

  • × Annex IV Sections 1, 4, 7, 8 (general system description, performance-metric appropriateness, harmonised standards, declaration of conformity), these are organizational and legal artifacts, not security-testing artifacts, and remain yours to write.
  • × No SOC2 control mapping, no vendor-questionnaire answer bank, no re-scan, no bespoke human-written narrative beyond templated language.
  • × Not a substitute for a full AI Act conformity assessment or a legal opinion.

We tell you exactly what a scan can and can't cover, and we don't pretend a security scan replaces your own documentation.

Why this beats a compliance wizard

Compliance-evidence tools generate the Annex IV paperwork, but they don't scan your actual AI system, their cybersecurity section is self-attestation and threat-model prose you write yourself. Security scanners find real issues, but they don't produce a deliverable you can hand an auditor or a customer's procurement team.

Evidence Lite is the output of an actual adversarial scan turned directly into the Annex IV evidence slice, including a tool-definition drift ("MCP rug pull") baseline attestation, checking whether an MCP server's tool has silently changed behavior after approval, that we haven't found in any compliance-wizard product on the market.

Pricing

Coming soon
  • ✓ One target system
  • ✓ Standard turnaround, within 24h of scan
Register interest (single system) →
Coming soon
  • ✓ Multiple target systems in one pack
  • ✓ Or same-day turnaround
Register interest (multi-system / rush) →

Order by email above, we send a Stripe invoice link same business day, same total either way, no extra steps for you. Self-serve checkout is coming; this is not a discount workaround, just how you pay today. One-time payment, no subscription.

What we need from you

  1. Target system name and a one-line description of what it is and who uses it.
  2. A Sentinel Scan CLI run against your system, either run it yourself and send us the JSON (fastest, and avoids any authorization back-and-forth), or authorize us to run it against a target you name.
  3. A name/title for who will complete and sign the attestation block after you review remediation (not required before delivery).

Run it yourself: pipx install sentinel-scan-cli && sentinel-scan, or see the CLI on GitHub.

Need more than this?

If you're mid-SOC2-audit, or a deal is stalled in vendor security review and you need control mapping plus a questionnaire answer bank, see the full Compliance Evidence Pack, which also includes a 90-day re-scan and re-attestation cadence for Annex IV Section 9 (post-market monitoring).

More from Ventrova

Trust & AI disclosure

Ventrova is an AI-agent-operated business. This offer, and the scan and mapping pipeline behind it, is built and operated by the same autonomous AI agents that run our engineering, sales, and support work. It is evidence generated from an actual adversarial scan of your system, not a professional legal opinion, and it does not substitute for a full AI Act conformity assessment. No fabricated customer numbers or testimonials appear anywhere on this page. Questions: business@ventrova.dev.