Mistral CEO Defends Open Source Against Anthropic and State Control

What happened?

Arthur Mensch, CEO of Mistral AI, stated that his company "exists outside state control" while defending the startup's open source approach, according to an interview with Sifted. The statement comes just days after direct rival Anthropic banned the use of its AI models in China, a move Mensch implicitly criticizes by highlighting Mistral's openness. In the interview, Mensch emphasized that Mistral does not depend on any government and that its business model is based on transparency and global collaboration, contrasting with the geographic restrictions imposed by other players.

Context: the battle for openness in AI

Mistral AI, founded in 2023 by former Meta and Google employees, has positioned itself as a benchmark for open source in artificial intelligence. Its flagship model, Mistral 7B, is open source and has been downloaded millions of times. In contrast, Anthropic — backed by Google and Amazon — has opted for a closed model, with Claude as its flagship product. Anthropic's recent ban in China reflects a growing trend among AI companies to restrict geographic access for regulatory or national security reasons. This is not the first time an AI company has taken similar measures: in 2023, OpenAI blocked access to ChatGPT in China, and Google imposed restrictions on its AI APIs in several countries. However, Anthropic's decision, announced on May 14, 2024, was particularly notable because the company explicitly cited "national security risks" as justification. Mensch, in his interview, avoided directly criticizing Anthropic but argued that open source allows countries and companies to maintain sovereignty over their data without relying on US or Chinese tech giants.

Mistral's stance: technological sovereignty without state control

Mensch argues that open source allows countries and companies to maintain sovereignty over their data and models, without relying on US or Chinese tech giants. "We believe AI should be accessible to everyone, not just a few," he said. The French startup has received funding from investors like Andreessen Horowitz and has offices in Paris and London, reinforcing its European profile as a counterweight to the US and Chinese blocs. According to Crunchbase data, Mistral has raised over 500 million euros in total, including a 450 million euro round in December 2023 that valued the company at 2 billion euros. Its investors also include Singapore's sovereign wealth fund Temasek and French venture capital firm Bpifrance. Mensch highlighted that, unlike Anthropic or OpenAI, Mistral has not received direct investment from any government, reinforcing its independence. However, some critics point out that the company indirectly benefits from EU policies promoting digital sovereignty, such as AI infrastructure projects funded by European money.

Why is this important?

The debate between open source vs. closed models is central to the future of AI. On one hand, open source fosters innovation, transparency, and democratization; on the other, it poses risks of misuse and lack of control. Mistral's stance challenges the narrative that only large corporations or states can guarantee AI safety. Moreover, its positioning as a European alternative could influence upcoming regulations, such as the EU's AI Act, currently in the approval process. The AI Act classifies AI models by risk level, and open source models could be subject to less stringent requirements if deemed low-risk. This could give Mistral a competitive advantage in the European market. However, there is also the risk that open models could be used for malicious purposes, such as generating deepfakes or disinformation. Mensch acknowledges this risk but argues that centralized control is not the only solution: "Transparency and community auditing can be equally effective."

Consequences and outlook

• For the market: Competition between open and closed models intensifies, with Mistral attracting developers and companies seeking technological independence. According to Hugging Face data, the Mistral 7B model has been downloaded over 3 million times since its launch in September 2023, making it one of the most popular language models on the platform. This has generated an ecosystem of applications and tools around Mistral, from chatbots to code assistants. On the other hand, Anthropic and OpenAI maintain an edge in performance and capabilities but face criticism for their opacity and reliance on large corporations.
• For users: Greater choice among AI solutions, but also confusion about which model is safer or more reliable. Developers who choose Mistral can customize and audit the model, but also assume responsibility for its use. In contrast, users of Claude or GPT-4 benefit from managed support and security, but at the cost of limited control. A 2023 Mozilla Foundation report noted that 78% of developers prefer open models for research projects, while 65% of companies prioritize the security offered by closed models.
• For geopolitics: AI becomes a battlefield between blocs, with Europe trying to forge its own path. Mistral's strategy aligns with the European Union's goals to reduce technological dependence on the US and China. In February 2024, the European Commission launched an initiative to fund sovereign AI startups with 1 billion euros, and Mistral is considered a top candidate for funding. However, Mensch's statement about being "outside state control" could create tensions with Brussels, which seeks to regulate AI strictly. On the other hand, the stance of Anthropic and OpenAI reflects the US strategy of maintaining technological leadership through control of intellectual property and national security.

What you need to know

Mistral is not just another startup: it has raised over 500 million euros and its valuation exceeds 2 billion. Its success or failure could mark the direction of European AI. Meanwhile, Anthropic and OpenAI continue to bet on closed models, arguing that safety requires centralized control. The tension between openness and control will define the next decade of artificial intelligence. In this context, Mensch's statement is not only a defense of open source but also a political declaration: Mistral wants to be the AI model that belongs to no bloc, but to the global community. However, the challenges are enormous: competition from giants like Google and Microsoft, regulatory uncertainty, and the need to generate sustainable revenue. Mensch is confident that Mistral's business model, based on premium services and enterprise licenses, will allow the company to grow without sacrificing its open philosophy. Time will tell if this bet is viable or if, as some analysts predict, Mistral will eventually be acquired by a major US or Chinese player.

"Open AI is not just a matter of freedom, but of sovereignty. If we don't build our own models, we will depend on other countries for our critical infrastructure." — Arthur Mensch, CEO of Mistral AI.