Bertrand Tavernier, Chief Technical Officer for Thales’s Secure Communications and Information Systems business: “This partnership with the CEA’s AI teams will combine the power of their research with our work at Thales’s AI accelerator, which brings together the Group’s technological expertise and deep knowledge of the defence and security sectors. Our customers — governments, armed forces, critical infrastructure operators — need trusted, sovereign generative AI solutions for their critical missions.”

Alexandre Bounouh, Director of the CEA’s List Institute, specialising in smart digital systems: “This partnership builds on the long-standing collaboration between the CEA and Thales and extends it to the sensitive issue of generative AI, combining the expertise and excellence of the CEA’s research teams in AI safety and security with Thales's AI accelerator's strengths in the strategic domain of defence and security. It will support the CEA’s mission in safety, security and artificial intelligence with our partners and all institutional and industry stakeholders in this field.”

Use cases for the armed forces

Generative AI can be developed to accelerate OODA command loops (observe, orient, decide, act) and implemented across the entire critical decision chain: sensing and data gathering, data transmission and storage, data processing and decision support.

Generative AI will serve as a trusted smart assistant for users, enabling them to dialogue easily and efficiently with complex systems with the aim of facilitating and accelerating human decision-making and the tempo of operations. For intelligence gathering, for example, multimodal generative AI will make it possible to simultaneously extract, process, correlate and interpret different types of information from multiple sources — such as the web, social media and sensors in a theatre of operations — to generate summaries and accelerate the production of reliable reports.

Thales’s AI Lab and the CEA will also focus on interoperability within coalitions. To simplify communication between member states in the context of a joint operation, trusted generative AI will facilitate interaction between operators and complex systems by translating their intentions into a sequence of actions and translating technical terms into the languages of the various nations involved.

Trusted AI for mission-critical systems

Thales is a major player in trusted, cybersafe, transparent, explainable and ethical AI for armed forces, aircraft manufacturers and critical infrastructure operators. In 2023, the Group was Europe’s top patent applicant in the field of AI for mission-critical systems. It has more than 600 AI experts and hosts around 100 PhD students every year in AI at Thales’s AI accelerator in the field of research (AI Lab), systems (AI Factory for decision support systems) and sensors (AI Sensors for sonar, radar, radio and optronics). Thanks to Thales's AI accelerator and its network of world-class industry, start-up and academic partners, Thales is building AI into more than 100 of its products and services. Benefiting from the most advanced sensor and system technologies, these AI capabilities cover the full spectrum of needs in defence, aviation, space, cyber security and digital identity. Trusted AI is designed to meet the specific security and sovereignty needs of Thales’s customers. It brings greater efficiency to data analysis and decision support and speeds up the detection, identification and classification of objects of interest and target scenes, while taking account of specific constraints such as cybersecurity, embeddability and frugality in critical environments.

CEA, a key player in trusted AI for mission-critical systems

In the field of artificial intelligence, the CEA is uniquely positioned across the entire AI value chain, with skills ranging from algorithms and software development tools to computing architectures and microelectronics. The hardware and software solutions developed by the CEA with its academic and industry partners are aimed at applications where AI quality and performance are critical. They include sovereign applications for which the CEA is its own customer (cybersecurity, defence, energy) and industrial applications to which the CEA contributes its expertise.

The CEA has developed high-level expertise in AI technologies, enabling the agency to support its industry partners in the development of innovative solutions. At its CEA-List Institute specialising in smart digital systems, the CEA focuses its research on the development of high-performance, trusted and sovereign models.

Working with Thales, the CEA teams will be applying their expertise more specifically to develop LLMs (large language models) and VLMs (vision language models) for the sectors and use cases targeted by Thales.