EASA AI Concept Paper levels explained

EASA’s levels are about how much authority an AI-based system has, not about how “smart” the model is. Issue 2 is the published guidance. Proposed Issue 03 is a consultation draft. Keep those statuses separate.

Status as of 14 October 2026: Issue 2 is still the published guidance, publication date 06/03/2024. The library page for EASA Artificial Intelligence Concept Paper — Proposed Issue 3 is still titled Proposed Issue 3, with publication date 03/06/2026. Comments were due no later than 12 August 2026. We found no published Issue 03 on that library listing.

Issue 2 (published): Level 1 and Level 2

EASA published Issue 2 on 6 March 2024. It refines guidance for Level 1 AI applications (those enhancing human capabilities) and provides comprehensive guidance for Level 2 AI-based systems.

Level 1 is human augmentation or assistance: the system enhances what a person can see or decide.

Level 2 introduces human-AI teaming (HAT) and sets the stage for AI systems that automatically take decisions under human oversight.

Issue 2 covers Level 1 and Level 2 AI applications but not yet Level 3. The paper does not constitute final detailed guidance. Keep that sentence in view when a program asks for “the EASA AI standard”: Issue 2 is published guidance, and EASA itself says it is not final detailed guidance.

Dated FAA versus EASA milestones, including NPA 2025-07, sit on FAA and EASA AI certification timelines. This page is the level taxonomy, not that date table.

Proposed Issue 03 (consultation): Level 0 and Level 3 added

On 3 June 2026 EASA released Proposed Issue 03 for a 10-week consultation, with comments due no later than 12 August 2026. EASA calls this the final Concept Paper deliverable foreseen under the EASA Artificial Intelligence Roadmap 2.0.

The news item says Issue 02 explored Level 1 AI (‘human augmentation/assistance’) and Level 2 AI (‘human-AI cooperation and collaboration’), and that proposed Issue 03 explores Level 3 AI applications, corresponding to ‘advanced automation’, including operations where the human end user may be either remotely present, or not present. It also addresses additional AI techniques, including reinforcement learning and symbolic AI.

The following sub-level wording is from the proposed Issue 03 PDF, not from published Issue 2.

The draft draws the Level 2 / Level 3 line on authority: at Level 2B there is a partial release of authority … under full oversight and control of the end user; at Level 3 the system is given full authority to make and implement decisions, under remote monitoring of an end user (Level 3A AI) or without end user involvement (Level 3B AI).

The same draft says it is covering Level 0, Level 1, Level 2 and Level 3 AI applications, but not covering ‘autonomy’ or ‘autonomous systems’. That exclusion matters next to civil UAS and weapon-system pages: proposed Issue 03 is still an EASA concept-paper consultation, not proposed 14 CFR Part 108 and not DoD Directive 3000.09.

EASA calls proposed Issue 03 the final Concept Paper deliverable foreseen under the EASA Artificial Intelligence Roadmap 2.0. A “final Concept Paper deliverable” is not, on that news item, an adopted AMC or a detailed specification. NPA 2025-07 remains a separate rulemaking proposal; see the timelines article.

What this is not

Issue 2 does not constitute final detailed guidance. Proposed Issue 03 is a consultation text. Neither document is a US FAA rule. The FAA has said EASA’s work helped inform the FAA roadmap; that is influence, not adoption.

Analysis, not sourced fact: if you are classifying a function today, start with published Issue 2 (Level 1 vs Level 2). Use proposed Issue 03 only to see where EASA is heading on Level 0 and Level 3. The 12 August 2026 comment deadline has passed; as of 14 October 2026 the library page is still Proposed Issue 3, so do not treat the draft as adopted guidance.

The EASA AI Concept Paper explainer covers learning assurance and the W-shaped process. The glossary defines those terms. For data sets and the ODD, see Data assurance for ML in avionics. For how that sits next to DO-178C and DO-330, see How to map an AI model to DO-178C / DO-330 tool qualification.

Questions

Is Issue 2 final detailed guidance for Level 1 and Level 2?

No. The Issue 02 PDF says the paper does not constitute final detailed guidance and covers Level 1 and Level 2 AI applications but not yet Level 3.

Does proposed Issue 03 cover autonomy?

The draft says it is covering Level 0, Level 1, Level 2 and Level 3 AI applications, but not covering ‘autonomy’ or ‘autonomous systems’.

Did the FAA adopt EASA’s levels?

The FAA has said EASA’s work helped inform the FAA roadmap. That is influence, not adoption.