As discussions on the simplification of the AI Act continue, progress on framework documents and governance structures is advancing in parallel. In this context, representatives of EU institutions, bodies and agencies (EUIs) gathered in Brussels on 10 February 2026 for the third meeting of the AI Act Correspondents Network.
The meeting marked a further step towards the Act’s practical implementation within EUIs. Participants discussed the ongoing simplification efforts, upcoming guidance and compliance challenges. A dedicated workshop focused specifically on the use of high-risk AI systems in recruitment and the safeguards required in human resources processes. The exchanges reaffirmed participants’ shared commitment to innovation and fundamental rights protection.
I had the pleasure of opening the meeting by underlining that 2026 is a decisive year for the AI Act, with transparency obligations becoming applicable and the EDPS assuming its responsibilities as Market Surveillance Authority and Notified Body for AI systems of EUIs. I also referred to the ongoing discussions on the Digital Omnibus, stressing that simplification must not weaken fundamental rights protection. Looking ahead, we are excited to continue our pilot regulatory sandbox project as a shared learning exercise to prepare the ground for a potential fully-fledged framework in the future.
The keynote address was delivered by Gemma Galdón Clavell, CEO of Eticas.ai, who offered a candid and practice-oriented reflection on the realities of AI compliance. She reminded us that many of today’s challenges, such as profiling and bias, are not entirely new but have evolved in scale and complexity. Rather than focusing solely on “making organisations compliant” on paper, she emphasised the need to build capacity to understand and manage the social and technical dimensions of the AI systems amplifying the known challenges. This is foundational for genuinely safe and trustworthy AI.
Representing the European Commission’s AI Office, Yordanka Ivanova shared an update on the current phase of the AI Act’s implementation. She outlined the latest progress on guidance for general-purpose AI models and high-risk AI systems and the consolidation of the governance framework. In addition, she addressed the simplification discussions under the Digital Omnibus, clarifying their impact on the AI Act’s timeline and supervision.
The workshop on AI systems in human resources, led by Emanuele Parisini and Manuel García Sánchez from the EDPS AI Unit, brought the discussion firmly into practice through an interactive case study on AI-driven recruitment tools. Participants were invited to assess a fictional high-risk system used for screening job applications, reflecting on questions of classification, registration, human oversight, bias and transparency. The format encouraged open exchange and critical thinking, highlighting the practical dilemmas EUIs may face when deploying such AI systems. This concrete scenario underscored the importance of meaningful human control and risk management.
The session on compliance considerations for Generative AI - led by the EDPS AI Unit and supported by colleagues from the Commission - addressed the key legal and practical questions surrounding its use within EUIs. Milda Macenaite and Félicien Schiltz examined roles and responsibilities under the AI Act, links with security considerations and the importance of strong internal governance. The discussion underlined that, even where generative AI is not used in high-risk scenarios, careful oversight, transparency and continuous monitoring are essential to ensure compliant and trustworthy deployment.
Spyros Sarigiannidis from the European Commission also gave an update on the state of play of its own general-purpose AI tool, GPT@EC, highlighting its governance model, hybrid infrastructure and compliance safeguards.
In her closing remarks, Sonia Pérez Romero, Head of the EDPS AI Unit, thanked participants for their active engagement and the high quality of the discussions within the network. She outlined the next steps, including the selection of the successful project application under the pilot regulatory sandbox and the continuation of this initiative in the months ahead. She also announced the forthcoming publication of the EDPS Compass, setting out the AI Unit’s operational priorities and approach under the AI Act for the next two years.
Reflecting on the discussions, it is encouraging to see that the network has entered a new phase, one defined less by anticipation and more by concrete action. The exchanges throughout the day showed a growing confidence in addressing complex compliance questions and demonstrated the value of this forum for shared learning. Such cooperation is particularly important in light of the adjustments introduced by the Digital Omnibus. In this evolving landscape, the network remains central to strengthening institutional preparedness.
Our next meeting is scheduled for 13 October 2026. Until then, our efforts will continue through close collaboration, practical exchange and the ongoing coordination of expertise across EUIs.