European startups bet on military AI as doomsayers lose ground in Paris

Two of Europe’s largest startups, Mistral AI and Helsing, have announced a partnership to develop AI systems for military use.

So what? Defence tech could be a lifeline for Europe at a time when its traditional industries are reeling from tariffs. But Mistral’s strategic pivot exposes an anxiety that the continent is falling behind in AI. That fear has metastasised in recent weeks due to

• Stargate – a staggering $500 billion investment in AI infrastructure announced by Donald Trump and the US tech industry; and
• DeepSeek – the Chinese startup which claimed its cutting-edge chatbot had been built at a fraction of the price of OpenAI’s ChatGPT, igniting a bull market for Chinese tech stocks.

France responded this week by hosting an “AI Action Summit” in which President Macron appeared in a deepfake video, announced €109 billion in AI investment and implored companies to “buy European”. But the event’s goals of “inclusivity” and “sustainability” jarred with a political shift towards innovation as a priority. The US and UK both refused to sign the final declaration on AI safety.

Mysterious Mistral. Macron’s attempts to promote a European AI champion show promise.

• Mistral has secured $1.2 billion in funding and helped to slow a brain-drain of French talent towards Palo Alto.
• More recently, though, excitement about the startup has cooled. Mistral’s founding strategy – to build cheaper and more efficient LLMs than US rivals – looks in danger of being undercut by DeepSeek.

Is the Helsing tie-up a panic move? “Mistral has been working on defence stuff for a while but DeepSeek made it way more urgent for them to pivot, or at least show off publicly what they’d been doing behind closed doors,” a source close to the company tells Sifted.

On the defensive. Having already secured deals with France’s defense industry, Mistral is reportedly in talks with the British and German governments. Its partnership with Helsing, a Munich-based manufacturer of autonomous attack drones valued at €450 million, will focus on building vision-language-action models that allow defence platforms to “understand their environment, communicate with humans and speed up decision making”.

Mistral isn’t alone in seeing the commercial advantage in AI-defence. According to a report by the Nato Innovation Fund

• venture capital funding for European defence startups grew 24 per cent in 2024, to $5.2 billion, while Europe’s broader VC market declined;
• startups developing technologies for “awareness,understanding, and decision making” have seen VC funding quadruple since 2020 to a record $1 billion; and
• Munich emerged as Europe’s top hub for defence and security investment, followed by Oxford and Paris.

Algorithm arms race. The cordon sanitaire that used to exist between tech companies and weapons makers is dissolving, thanks not least to economic nationalism. In the last few months

• OpenAI announced a new partnership with the defence tech startup Anduril. Anduril and Palantir, a data analytics firm, are in talks to form a consortium to bid for US defence contracts;
• Meta has announced plans to make its AI models available to defence contractors Lockheed Martin and Booz Allen;
• Google ditched a longstanding ban on using AI to develop weapons and surveillance tools, seven years after it caved to staff pressure to drop “Project Maven” with the Pentagon.

Vibe shift. Europe is having regulation regrets. The AI doomsayers have lost ground while national governments, including the UK’s, have become fixated on the technology as an antidote to lacklustre growth.

What’s more… The die may be cast. In the words of Max Tegmark, president of the Future of Life Institute, the progress made since 2023 on AI safety has been “guillotined”.