Trust · EU AI Act and privacy

Responsible AI for infrastructure inspection.

RailEyes uses image analysis to help rail, road and city teams find visible defects and assets that need attention. Software proposes observations, a person reviews every finding, and the client decides what to do. This page summarises how the service is classified under the EU AI Act and how footage is protected.

Assessment of 16 November 2026, based on Regulation (EU) 2024/1689 as amended by the Digital Omnibus on AI (Regulation (EU) 2026/1744) and the Commission’s draft guidelines on high-risk classification of 19 May 2026.

EU AI Act

Not a high-risk AI system.

Prohibited practices (Article 5)

No. Faces are detected only to anonymise them. No person is identified or categorised, and no facial-recognition database is created.

High-risk: biometrics (Annex III, point 1)

No. Biometric identification means establishing who a person is. The Commission’s draft guidelines list AI systems whose sole purpose is anonymising personal data outside this use case.

High-risk: critical infrastructure and road traffic (Annex III, point 2)

No. RailEyes is not a safety component in the management or operation of road traffic: it delivers human-reviewed findings for maintenance planning, after the survey, and acts on nothing in real time. The draft guidelines place predictive-maintenance and analytics tools that support other safeguards outside this use case.

Transparency (Article 50)

No obligation arises: the service does not interact with people or present generated content as real. Blurring faces and number plates is standard editing.

AI literacy (Article 4)

Applies, and is in place: RailEyes trains its operators and reviewers in the system’s use and limits.

RailEyes reviews this assessment before any change of intended use, such as real-time alerts that trigger traffic-control actions, integration into a traffic-management system, or automated decisions without human review. Obligations for Annex III high-risk systems apply from 2 December 2027 under the Digital Omnibus.

Privacy by design

Footage protected from the first frame.

Anonymised at capture

Faces and number plates are anonymised on the vehicle as frames arrive. Raw frames exist only in volatile memory and are never written to storage, including after a crash or restart.

A second check

A secondary detector screens every frame before footage can be saved or sent. Frames that fail are reprocessed or discarded.

Processing in France

Anonymised footage travels over an encrypted, EU-hosted VPN to RailEyes-operated servers in France. No cloud service holds programme footage.

People in the loop

Every finding is checked by a reviewer before it reaches a client, and every image is cleared for release. Clients decide what to do through their own procedures.

Roles and retention agreed per project

Who is controller and processor, the legal basis, public notices and retention periods are recorded in each project’s documents before any recording starts.

Documents and contact

The Smart Cities Programme privacy notice is open to everyone. The full AI Act assessment, the technical and organisational measures and the DPIA template are in the programme data room. Privacy questions: privacy@raileyes.eu.