Research Engineer In ai Security For Specifying And Implementing Regulatory Compliance Tests For The eu ai Act H/F INRIA

Research engineer in AI security for specifying and implementing regulatory compliance tests for the EU AI Act
Le descriptif de l'offre ci-dessous est en Anglais
Type de contrat : CDD

Contrat renouvelable : Oui

Niveau de diplôme exigé : Thèse ou équivalent

Fonction : Ingénieur scientifique contractuel

A propos du centre ou de la direction fonctionnelle

The Inria research centre in Lyon is the 9th Inria research centre, formally created in January 2022. It brings together approximately 300people in 19 research teams and research support services.

Its staff are distributed at this stage on 2 campuses: in Villeurbanne La Doua (Centre / INSA Lyon / UCBL) on the one hand, and Lyon Gerland (ENS de Lyon) on the other.

The Lyon centre is active in the fields of software, distributed and high-performance computing, embedded systems, quantum computing and privacy in the digital world, but also in digital health and computational biology.

Contexte et atouts du poste

As part of the implementation of the EU AI Act, Inria coordinates the NoLeFa-84 project (see https://nolefa.eu), aimed at prototyping a new facility for testing and technical support to the EU authorities in charge of verifying the proper application of the regulation by AI system providers.

This includes for instance technical verifications related to data quality obligations, performance, robustness and AI security criteria, or requirements on transparency of the AI system.

Article 15 of the EU AI Act includes specific obligations regarding AI security, with a particular focus on attacks such as adversarial attacks, data or model poisoning, and confidentiality attacks. The exact technical criteria for compliance to these obligations remain however unspecified by the regulation itself, and are left to be further clarified and detailed by technical standards developed by the European technical committee for AI standardisation (CEN-CENELEC JTC 21), under a mandate from the European Commission.

Given the fast-evolving nature of this field, the objective of this standardisation work is to identify the most meaningful criteria and thresholds that can be used by AI system developers to determine the defense methods that are legally suitable (e.g. the percentage of attacks that an adversarial detection method can detect), rather than setting in stone one particular defense method or imposing too challenging expectations on providers.

Mission confiée

The goal of this position is to contribute concurrently to the development of CEN-CENELEC JTC 21's technical specifications detailing the legal obligations of the EU AI Act into practical criteria for developers, and to the implementation of concrete testing methods to verify them on the regulated AI systems.

The person will be co-supervised by the technical coordinator of the NoLeFa-84 project and by a researcher specialized in AI security (Antoine Boutet, PRIVATICS team). The work will be conducted in close coordination with Inria's European partners in the NoLeFa-84 consortium (comprising another public body, 3 SMEs/startups and a research labs association), including standardisation specialists and R&D engineers. Synergies will also be sought with the PANAME project (see https://files.inria.fr/ipop/2025-06-26-paname/) related to the privacy auditing of AI models, in the frame of the IPoP project (see https://files.inria.fr/ipop/).

Following these contributions to the NoLeFa-84 project, opportunities can be envisioned for pursuing with a new contract on similar activities at INESIA, the French National Institute for Evaluation and Security of AI, that Inria is a founding member of.

Principales activités

- State of the art on attack/defense methods in AI security
- Writing technical specifications and criteria for compliance to the EU AI Act
- Participation in the working groups of the European technical committee for AI standardisation (CEN-CENELEC JTC 21), for contributing to the EU AI Act implementation standards
- Implementation of testing tools for AI systems to check their compliance to legal obligations on AI security from the EU AI Act
- Participation in the IPoP working groups related to the PANAME project

Compétences

- PhD in AI security, with advanced knowledge in at least 2 attack categories (and associated defense methods) among: adversarial attacks, data or model poisoning, confidentiality attacks
- Fluency in English: ability to participate in complex technical discussions and argue to defend technical positions
- Interest for the EU AI Act regulation and for bridging technical and legal considerations

Avantages

- Subsidized meals
- Partial reimbursement of public transport costs
- Leave: 7 weeks of annual leave + 10 extra days off due to RTT (statutory reduction in working hours) + possibility of exceptional leave (sick children, moving home, etc.)
- Possibility of teleworking (after 6 months of employment) and flexible organization of working hours
- Professional equipment available (videoconferencing, loan of computer equipment, etc.)
- Social, cultural and sports events and activities
- Access to vocational training
- Social security coverage under conditions

Rémunération

From 2,692 € (depending on experience and qualifications).