By the end of the year, the European Commission will launch three flagship projects under Horizon Europe to support the development, training and testing of artificial intelligence systems for cyber threat detection, incident response and self-healing capabilities. It will also support the development of “sovereign European AI cyber capabilities, including for defence, through the European Competitiveness Fund”, under the next Multiannual Financial Framework (MFF), “ensuring that Europe is better equipped to protect its strategic interests and reduce critical dependencies”. These initiatives form part of the Action Plan on Cybersecurity and Artificial Intelligence presented by the Commission on Tuesday 7 July.
As Agence Europe had already reported (see EUROPE 13902/7), the Action Plan provides for a series of concrete measures, including the establishment of a Union capacity to evaluate artificial intelligence models, particularly from a cybersecurity perspective, before they are placed on the market – a capacity that should be operational in 2027. The Commission’s Executive Vice-President, Henna Virkkunen, explained that the institution will launch a call for tenders in the autumn and define the criteria for determining the types of entities that will be able to take part. “Research institutes and various national cybersecurity agencies are expected”, she explained, without wishing to rule out non-European entities, while specifying that this was a “very sensitive” sector.
The Plan also provides for the development of a European framework designed to ensure a structured approach to companies’ access to AI models. However, Ms Virkkunen acknowledged that this framework “will not impose legal or binding obligations” on providers of AI models. Instead, it will aim to facilitate guidance on access criteria and conditions, and improve transparency concerning the gradual placing on the market of models and verification procedures, and could serve as a basis for cooperation with international partners.
The Plan also provides for the creation of a secure platform for testing AI models in cybersecurity and for ensuring the safe use of AI models in critical sectors such as finance, energy, health, transport and public administrations.
It also aims to accelerate the development of European AI-based cybersecurity capabilities and to help SMEs deploy existing AI tools in order to strengthen their resilience to cyberthreats. “Our immediate priority is clear: identify (...) the most critical vulnerabilities faster”, including in open source digital infrastructures, Ms Virkkunen said, calling on Member States to implement EU rules such as the NIS2 Directive and the DORA and Cyber Resilience Acts as swiftly as possible.
Hundreds of billions of euros in investment needed in AI. “Developing sovereign frontier capabilities will entail hundreds of billions of euro investment needs which can only be partly covered by public finance”, the Commission stresses in its document.
“To develop our own sovereign capacities in the field of cutting-edge AI models, we must mobilise considerable additional private investment”, Henna Virkkunen warned when presenting the Action Plan during a plenary sitting of the European Parliament. She stressed that the new ‘Tech Facility’, currently being examined under the Tech Sovereignty Package (see EUROPE 13880/1), “could be a game changer” for investing in “European strategic companies”. She also called for the Savings and Investments Union to be established more quickly, which she described as an “absolutely essential” element for financing European technological sovereignty.
By the end of the current Multiannual Financial Framework, under the Horizon Europe and Digital Europe programmes, €200 million is expected to be mobilised for research and development into advanced European AI-based cybersecurity technologies, as well as into the cybersecurity of artificial intelligence itself.
At the same time, the Union is investing, through the European Innovation Council (EIC), in specialised European high-tech companies, with an envelope of €100 million intended for start-ups and scale-ups developing strategic defence technologies by the end of 2026, before continuing similar support in 2027. The Commission also announced that the ‘Scaleup Europe Fund’ will make its first investments in strategic growth-stage companies in 2026.
Political groups in the European Parliament call for European solutions and more investment in innovation. “We will not be able to turn away from American suppliers overnight”, warned Aura Salla (EPP, Finnish). She stressed the need to reduce risks weighing on supply chains and to develop European solutions through strategic public procurement, and an ambitious industrial policy. “We cannot ignore this reality”, added Alex Agius Saliba (S&D, Maltese), stressing that the more economies and societies become digitalised, the more vulnerable they become to cyberattacks. In his view, AI systems are rapidly becoming more powerful and Europe must strengthen its own capabilities.
The PfE and ECR groups also called for greater support for European technology companies and innovation. “The United States innovates, the European Union regulates”, criticised Rada Laykova (ESN, Bulgarian), taking the view that all Member States should be concerned by the United States’ decision not to include the European Union among its “trusted partners” authorised to access the latest cutting-edge AI models developed by Anthropic.
Bart Groothuis (Renew Europe, Dutch) warned companies that they had, “at best, only a few months to fix vulnerabilities”. Believing that Chinese models would arrive on the market “soon”, he stressed the need to develop cutting-edge AI models “in Europe”.
“Europe was sleepwalking into a new dependency”, warned Markéta Gregorová (the Greens/EFA group, Czech). She regretted that the Action Plan contained no mechanism capable of preventing the Union from once again being deprived of access to cutting-edge AI models and argued in favour of the emergence of “European champions”. Finally, João Oliveira (The Left, Portuguese) said that States’ responsibilities in terms of control must be clearly defined. (Original version in French by Ana Pisonero Hernández)