Home / The Act / Chapter III / Article 15

Article 15 — Accuracy, robustness and cybersecurity

APPLIES DEC 2, 2027 WAS 2026-08-02 AMENDED BY OMNIBUS

Chapter III · High-risk tier
Source: EUR-Lex · CELEX 32024R1689
Verified 2026-07-16

PLAIN-LANGUAGE SUMMARY — ORIGINAL INTERPRETATION, NOT THE OFFICIAL TEXT

High-risk AI systems must be engineered to reach a suitable level of accuracy, stay reliable, and resist tampering throughout their working life. Declared accuracy figures belong in the instructions for use, and robustness can rest on redundancy or fail-safe designs. Systems that keep learning after release must limit biased feedback loops. Protective measures should match the actual risk, countering threats such as data poisoning, adversarial inputs, and attempts to exploit model flaws.

AMENDED BY THE DIGITAL OMNIBUS ON AI

This provision's application date moved under the Omnibus. It now applies from 2027-12-02 (originally 2026-08-02). Source: Omnibus final text (OJ pending) · verified 2026-07-16

OFFICIAL TEXT Read verbatim on EUR-Lex ↗

The authoritative wording lives on EUR-Lex, quoted there in all 24 languages. We deep-link to the paragraph level rather than reproducing the full text. The consolidated post-Omnibus version will be linked the day its CELEX publishes.

AT A GLANCE

Chapter: III — High-risk AI systems

Applies: Dec 2, 2027

Tier: High-risk tier

High-risk obligations →Harmonised standards →High-risk classification →

PENALTY EXPOSURE

€15M or 3%

of worldwide turnover — see Art. 99 penalties framework.

← PREVIOUS · ART. 14 Human oversight NEXT · ART. 16 → Obligations of providers of high-risk AI systems

ELSEWHERE IN CHAPTER III — High-risk AI systems

Art. 6 Classification rules for high-risk AI systems Art. 7 Amendments to Annex III Art. 8 Compliance with the requirements Art. 9 Risk management system Art. 10 Data and data governance Art. 11 Technical documentation Art. 12 Record-keeping Art. 13 Transparency and provision of information to d