Mistral versus the prophets of the AI apocalypse

Arthur Mensch, CEO of the French firm Mistral AI – which he himself has described as “the only fully European AI company, from computation right through to deployment in businesses” – accused American laboratories in an interview published yesterday in Le Monde of exploiting fears about the risks of artificial intelligence to “control and shut down the market for their own benefit”. According to Mensch, the major laboratories are painting apocalyptic scenarios to convince governments that only a few companies – including themselves – can provide safe models – the so-called ‘regulatory capture’.

After some GPT agents escaped OpenAI’s control this summer, breaching Hugging Face (a global community of programmers), Dario Amodei (head of Anthropic) called for a slowdown in the pace of AI development, and OpenAI CEO Sam Altman agreed. However, David Sacks, head of the White House’s scientific advisers, responded to the proposal on X on 13 September: “Go ahead. You have a duopoly and you don’t need permission. But stop pretending the reason is altruism: the market already punishes models that behave in unpredictable or unauthorised ways. Quite simply, it is now in your best interests to trade power for reliability and predictability.” On 23 September, speaking before the UN Security Council, the two CEOs reiterated their call for international standards for the most powerful models and a common system for reporting incidents, although Altman has now stated that he is opposed to anti-competitive measures.

Mensch, speaking to Le Monde, did not deny that risks exist, acknowledging the need for more sophisticated controls and stating that those who use AI must monitor it and be held criminally liable. However, he took issue with those who, in his view, are actually trying to squeeze out smaller competitors and open-source models (which Mistral has focused on), often favoured by businesses because they are customisable, independent of suppliers and more cost-effective.