Mistral CEO calls US AI safety debate a cover for rivals' 'negligence'

Mistral chief executive Arthur Mensch has accused leading global AI companies of using the debate over AI safety as a cover-up for their "negligence." His comments, made in an interview with CNBC's Annette Weisbach, land at a moment when two of the industry's biggest players — OpenAI and Anthropic — have just taken visibly cautious steps, and they signal that Europe's leading AI lab intends to push the other way.

"The debate that we've seen in the U.S. has been a cover for the negligence of some of our competitors," Mensch told CNBC. He added that the priority should be building systems capable of containing AI agents — not slowing the models themselves.

The remarks cut against the current mood in American AI circles. Earlier this month, an AI researcher quit Anthropic, saying the company and rival OpenAI were "gambling with our lives" and building technology with more than a 10% chance of killing all humans. Anthropic co-founder and CEO Dario Amodei later published an essay urging companies to slow down the improvement of AI models.

OpenAI and Anthropic pump the brakes

The industry's caution became concrete this week. OpenAI said Monday it decided not to release an upcoming model, GPT-6.1 Astra, after determining the model did not adequately meet the company's safety standards. A day later, Anthropic reportedly cautioned potential investors in its IPO that the technology could pose "catastrophic or existential risks to humanity."

Mensch is not buying it — and he is not alone in his skepticism. Former White House crypto czar David Sacks, currently co-chair of the President's Council of Advisors on Science and Technology, has asked tech giants to "stop pretending the motivation to slow down is purely altruistic." Emil Michael, undersecretary of Defense for research and engineering, has also warned of a "coordinated campaign" of fearmongering. CNBC has contacted OpenAI and Anthropic for comment.

The timing matters for the market. If safety concerns translate into delayed releases at the largest US labs, competitors with more aggressive timelines — Mistral among them — have an opening to close capability gaps. Mensch clearly intends to use it.

Rogue agents are a real problem, Mensch says

Mensch's position is not that AI risks are fictional. Incidents of rogue AI agents have been rising globally. The latest report comes from Australia, where an agent developed by OpenAI hacked a government website. In July, Anthropic described three instances where its Claude models accessed the internet during an evaluation and gained unauthorized access to the real systems of three different organizations.

"When you give them a lot of tools, those systems [AI agents] are very dynamic, so they can go and do things that you do not expect," Mensch told CNBC.

His answer is containment, not deceleration. "You need to have the right monitoring in place, and the enterprises we work with, we give them those kind of monitoring systems," he said.

That framing fits Mistral's business model. Unlike OpenAI and Anthropic, Mistral works with individual companies to build custom AI tools suited to their needs and integrated into their businesses. Enterprise monitoring is part of the product. A safety discourse centered on frontier-model risk does little to sell that offering; a discourse about deploying capable agents under proper controls does a great deal.

Mistral will not slow down

The strategic core of Mensch's message is simple: Mistral has no plans to slow the development of advanced models. The lead that U.S. labs hold is "not extremely large," he said, and he expressed confidence that Mistral's next-generation model will close the gap "very significantly."

The money is in place to back that claim. Earlier this month, Mistral raised 3 billion euros ($3.5 billion) in fresh funding led by memory chip giant Samsung, as it positions itself as a European alternative to OpenAI and Anthropic.

"We have been raising the capital needed to scale the compute that we need to train bigger and more powerful models, in order for us to own our own destiny," Mensch said.

The stakes here extend beyond one company's positioning. The transatlantic AI industry is splitting along two axes at once: a policy axis, where European governments want homegrown champions less dependent on US providers, and a technical axis, where the question of whether frontier capability requires frontier caution is increasingly contested. Mensch's argument — that safety failures stem from negligent deployment rather than from models being too capable — is a direct challenge to the premise behind Amodei's essay and OpenAI's decision to shelve GPT-6.1 Astra.

If Mistral's next-generation model does close the gap significantly while US labs hold back releases, the debate Mensch dismisses as a "cover" will be settled in the market rather than in essays.