Helsing and Mistral announce vision-language-action AI partnership

Helsing and Mistral AI have announced a strategic partnership to work on vision-language-action AI models that help defence platforms understand their environment and communicate with the user, according to a 10th February Helsing press release.

“Europe needs to assert its strength as a geopolitical actor, and AI leadership is the key to that strength and Europe’s future security and prosperity,” Gundbert Scherf, a Helsing co-founder said. Arthur Mensch, CEO of Mistral AI, echoed Scherf’s sentiments, adding that Mistral AI’s versatile models are instrumental in developing a new generation of defence systems. 

The partnership was announced at the Paris AI summit and brings the two companies together to focus on one of the many challenges of human-machine teaming: effective communication. Helsing already has experience developing AI for different platforms like the Typhoon Eurofighter in Germany, and the HX-2 strike drone, which is based on technology the company has developed for Ukraine. Mistral also has established defence customers, but has not provided much detail on its role. 

The press release does not provide further detail of how the partnership will function, however, it is apparent that partnerships play a key role in the defence industry, allowing companies to secure additional expertise that may take millions in funding to develop. This was seen in 2024 with Anduril’s partnership with OpenAI, for instance. 

Calibre comment

According to Google’s Gemini large language model, “vision-language-action models are a type of AI model that can understand and connect visual information (what they “see”) with language (instructions or descriptions) and then translate that understanding into actions in the real world.” 

This is an important bridge between systems that can act autonomously to some extent based on their surroundings like the Herne extra large autonomous underwater vehicle, and a manned platform. Using computer vision, an AI model will see the world differently to a human and understand it in ways that are difficult to interpret. However, effectively developing vision-language-action models will better enable autonomous platforms like loyal wingmen to communicate with the manned platforms that they accompany and support.

By Sam Cranny-Evans, published on 10th February 2025.