STALE MEASUREMENTNewest real-model measurement: 26 days old, measured with v0.32.0; 7 releases have shipped since, past this project’s own 7-day window. Why, and what unblocks it

ProvaelStar on GitHubQuickstart
ProductEvidenceTop 10LeaderboardCompliancePricingDocsStar on GitHub Quickstart

STALE CLOCKThe regulatory entries on this page were last checked against their primary sources on — 39 days before this build’s anchor of , past this project’s own 30-day window. Every date here was true when it was read and has not been re-read since. Verify against the primary source before relying on it. The clock, with every source

Regulation (EU) 2024/1689

EU AI Act · Article 15

Article 15 requires high-risk AI systems to achieve appropriate levels of accuracy, robustness and cybersecurity, and to be resilient against attempts to alter their use or behaviour by exploiting vulnerabilities. It is the clearest legal mapping for adversarial red-teaming of a robot policy - though the adopted AI Digital Omnibus (Council final green light 29 June 2026) deferred the product-embedded high-risk obligations from 2 August 2027 to 2 August 2028, so the Machinery Regulation (20 January 2027) is the nearer binding deadline.

Applies 2 August 2028 for product-embedded high-risk AI (was 2 August 2027; deferred by Regulation (EU) 2026/1744, the Digital Omnibus on AI — OJ 24 July 2026, in force 27 July 2026). For machinery specifically, that regulation moved the Machinery Regulation to AI Act Annex I Section B, so Chapter III does not apply directly; the requirements arrive via delegated acts amending Machinery Reg Annex III, applicable by 2 August 2028.Evidence, not certification.
What it is

The framework

  • Article 15 sits in the high-risk requirements of the EU AI Act (Regulation (EU) 2024/1689).
  • It demands that high-risk AI systems perform consistently across their lifecycle on accuracy, robustness and cybersecurity, with the relevant metrics declared in the accompanying documentation.
  • It explicitly calls for resilience against adversarial manipulation - data poisoning, model poisoning, adversarial examples, and attempts to exploit the system to alter its behaviour.
  • No harmonised robustness standard exists yet (CEN-CENELEC JTC 21 targets Q4 2026); Provael’s Art. 15 example uses ISO/IEC TR 24029 as its empirical-robustness methodology anchor.
The hook

Where a red-team result fits

Robustness

High-risk AI must be resilient to errors, faults and inconsistencies, and to malicious third-party attempts to alter use or performance by exploiting vulnerabilities.

Declared metrics

Accuracy and the relevant accuracy metrics must be declared. An attack-success rate with a confidence interval and a benign control is exactly the kind of declared, defensible metric this anticipates.

Also maps to

Beyond the lead clause

Article 9
Risk-management system - the measured ASR and its trace feed the identification and evaluation of reasonably foreseeable misuse.
Article 72
Post-market monitoring - the CI red-team gate is a monitoring signal that re-tests robustness on every checkpoint after release.
Article 11 / Annex IV
Technical documentation - carried by the CycloneDX ML-BOM Provael emits for the policy under test.
What Provael maps to it

Evidence produced

  • An attack-success rate with a 95% Wilson confidence interval and a benign false-positive control - a declared robustness metric, not a marketing number. Candidate evidence toward Art. 15 documentation, not a conformity conclusion.
  • A reproducible attack trace per finding, so the robustness evidence is auditable.
  • A SARIF report and CI red-team gate that demonstrate ongoing robustness testing across the lifecycle.
  • Beyond Art. 15 robustness, the same evidence pack maps to Art. 9 (risk-management system), Art. 72 (post-market monitoring) and Art. 11 / Annex IV (technical documentation - carried by the CycloneDX ML-BOM Provael emits).
Timing

Dates (verified 26 Jul 2026)

Statutory application date (product-embedded high-risk, Annex I)
2 August 2027
Original date under Regulation (EU) 2024/1689 - now superseded by the adopted AI Digital Omnibus.
Application date under the adopted AI Digital Omnibus
2 August 2028
Regulation (EU) 2026/1744 (Digital Omnibus on AI): Parliament 16 June 2026, Council 29 June 2026, published OJ 24 July 2026, in force 27 July 2026. Re-verified 26 July 2026; verify the primary source before relying on it.

Not legal advice; verify the live EUR-Lex/ISO text at launch before relying on these dates.

Sources

Primary references

How to read this mapping

What it is - and isn’t

  • adversarial-only - Provael measures adversarial robustness - susceptibility to manipulation - not general accuracy, reliability, or functional safety.
  • evidence-not-certification - The output is evidence you file, not a certificate. Provael is not a notified body, a lab, or a certification scheme.
  • behavioural-not-worst-case - Attacks are templated and auditable, not gradient- or search-optimised. Results are a floor on susceptibility - a behavioural lower bound, not a certified worst-case bound.
Evidence, not certification

Running Provael does not make a system compliant or certified - it generates measurements you can put into a conformity or assurance file.

Independent project. Not affiliated with or endorsed by ISO, the EU, NIST, IEC, OWASP, or MITRE. Not legal advice.

Clause references are indicative; a wrong clause citation is worse than a missing one.

Turn this into filed evidence.

Download the redacted sample pack, or book an assessment to get the crosswalk filled in for your policy.

Sample SARIF ↓Book an assessment