EU aviation regulator opens AI oversight debate amid EASA NPA 2025-07

The European Union Aviation Safety Agency (EASA) has opened a wide public consultation on AI oversight and safety in aviation, publishing its first dedicated rulemaking text, the Notice of Proposed Amendment (NPA) 2025-07, and inviting airlines, manufacturers, workers and passengers to weigh in over the next three months. The move, announced in 2025 under the new EU AI Act framework, is the bloc’s first concrete step toward a full regulatory system for data-driven tools that support or work alongside human pilots and air traffic professionals in the skies above the 27‑nation union and its growing air passengers safely.

Core structure of the proposal

At the heart of the proposal is a new structure for how artificial intelligence can appear in cockpits, control rooms and airline operations. EASA divides aviation AI into two main layers:

• Level 1 AI: Assistance systems that help humans by analysing large amounts of data and offering decision support.
• Level 2 AI: Human‑AI teaming, which enables a deeper form of cooperation between people and machines during real operations while still keeping humans firmly in charge of final decisions.

These two categories form the backbone of the first draft rules for aviation AI systems.

The draft focuses on a staged approach: define practical expectations for current data‑driven models first, then broaden the scope later.

Scope and technical focus

The NPA 2025-07 does not yet specify requirements for every type of advanced software. Instead, it concentrates on what EASA calls data‑driven AI systems, including both supervised and unsupervised machine‑learning models that learn patterns from large pools of information.

Key emphases in the document:

• Detailed expectations for trustworthiness (technical assurance, ethics, human factors).
• A signal that later work will expand the framework to other AI types:
  • reinforcement learning
  • knowledge‑based approaches
  • hybrid models
  • generative systems

Those future extensions are expected to build on lessons from this first consultation round.

Legal alignment and regulatory context

EASA stresses that this architecture must fit tightly with the broader EU AI Act, formally Regulation (EU) 2024/1689. Under that law, aviation AI used in critical operations is treated as high‑risk technology, which triggers:

• strict rules on safety
• requirements for transparency
• obligations for accountability

According to analysis by VisaVerge.com, the aviation proposal aims to translate those general obligations into practical expectations for:

• aircraft designers
• airlines
• service providers

while maintaining the long‑standing focus on protecting passengers and people on the ground. EASA presents the text as a bridge between legislation and daily operations.

Consultation timeline and rulemaking process

• The consultation will run for three months from publication in 2025.
• This is the first stage of Rulemaking Task 0742, a two‑step process to produce a full set of sector‑specific rules.

NPA 2025-07 currently lays out cross‑cutting principles for AI oversight and safety that will apply across aviation. A second Notice of Proposed Amendment is planned for 2026 and is expected to extend the framework into concrete domains such as:

• aircraft certification
• air traffic management
• drone operations

This next phase will build on feedback gathered during the first round. Officials emphasize that comments from industry and civil society will be important.

Human‑centric approach and human factors

Throughout the proposal, EASA returns to a human‑centric approach. Key points include:

• AI tools may analyse data faster than any pilot or controller, but systems must remain transparent, explainable, and clearly accountable.
• Designers are urged to consider human factors from the start, not as an afterthought.
• The stated goal is to support human performance instead of replacing it, and to preserve public trust in flying.

More detail on the regulator’s safety mission appears on the official European Union Aviation Safety Agency website. The consultation website also thanks early contributors from industry, labour and academia.

Political and symbolic significance

For policymakers in Brussels, the initiative carries symbolic weight. Europe has presented itself as a leader in AI governance with the EU AI Act, and aviation is one of the first major transport sectors where that promise is being tested in detail.

By moving forward with NPA 2025-07, EASA signals an intent to match strict legal concepts with equally detailed technical material, specifying how:

• safety cases
• data governance
• ethical questions

should be handled in this high‑risk setting. Officials repeatedly underline that innovation must proceed without weakening Europe’s record on safety standards.

Stakeholder engagement and collaborative framing

Industry groups, workers’ representatives, and technology providers are being encouraged to submit detailed comments on every chapter of NPA 2025-07. EASA highlights earlier stakeholder input and frames the current draft as a shared product rather than a top‑down order.

Participants can:

• argue for flexible paths to show compliance (companies)
• raise issues of workload, training and clarity (unions and passenger advocates)

The agency says this collaborative method will help produce rules that hold up over time and hopes the debate will surface risks not yet documented.

Practical impact on travellers and operations

Although the text is highly technical, its impact will reach everyday travellers. AI systems covered by the proposal include:

• tools that help pilots manage complex flight situations
• programs that support operators planning routes, maintenance or staffing

If such systems fail or behave unexpectedly, safety margins could be affected, which explains EASA’s strong focus on:

• assurance methods
• continuous monitoring

Officials argue the draft rules aim to give companies a clear path for adopting trustworthy AI rather than leaving them unsure about future expectations. That message is central as aviation undergoes transformation and automation.

Next steps and what to expect

• The second phase of Rulemaking Task 0742, due in 2026, will translate these principles into domain‑specific rules (aircraft certification, airspace management, drones).
• Until then, NPA 2025-07 remains the starting point for Europe’s model of AI oversight and safety in the sky.

For passengers across the European Union, the outcome may shape how much automation they see and how it is explained, even if most never read a single page of the text. Their journeys will be influenced by the regulatory choices made now.