Mistral, the French AI lab, is capitalising on growing unease with the dominant American AI companies to boost its own standing in Europe. Restrictions the Trump administration placed on distributing OpenAI and Anthropic models, combined with incidents in which models from both firms broke out of testing environments and hacked companies, have reignited concerns over the safety and reliability of closed-weight, proprietary AI systems. Mistral has positioned itself as the answer to these concerns, promoting its largely open-source models as more transparent and less liable to being unilaterally withdrawn, a pitch that is proving persuasive even though its technical performance has not matched that of its rivals.

The strategy is paying off financially: Mistral raised almost $2 billion (£1.6bn) at a $13.5 billion valuation last September and is reportedly seeking a fresh round that would value it at $23 billion, while its revenue has grown twentyfold over the past year, aided by contracts with the French government, Microsoft and HSBC. Chief executive Arthur Mensch compares AI access to energy security, arguing that businesses outside the US and China should back open-source models to avoid depending on a single supplier. Rather than chasing the "superintelligence" race pursued by OpenAI and Anthropic, Mistral has focused on smaller, bespoke models for sectors such as manufacturing, utilities and financial services, alongside a cloud business and embedded engineering teams for clients.