AI News Today October 6 - Pentagon Cuts Off Anthropic, Europe's 'Le Chonk' AI Challenges China

If you invest through a retirement account, the next big stock market listing may be an AI company the Pentagon has just cut loose.

At the same time, France's Mistral has launched an AI model it says beats Chinese rivals on cyber defence, and it plans to let anyone download it.

Here's everything you need to know today.

Why Did the Pentagon Stop Using Anthropic's Claude?

Weeks before its expected stock market debut, Anthropic has been cut off by the US military.

A Pentagon official told the BBC the department has stopped using Anthropic's tools, months after Defence Secretary Pete Hegseth branded the company a supply chain risk for refusing to drop safeguards against mass surveillance and autonomous weapons.

Anthropic has sued to overturn that label. Sources say Claude was still in use last week, including in operations against Iran, through Palantir's Maven system.

One analyst says that shows how hard it is to pull an AI out once it is built into a system.

For anyone considering Anthropic's IPO, this is a fresh risk, because standing firm on safety has just cost it a major government customer. The Pentagon has made its choice, and in AI, a company's stance on safety now carries a price tag.

What Is Mistral Large 4, the "Le Chonk" AI Taking On China?

Europe has just made its boldest AI claim yet.

Mistral calls its new Le Chonk model, officially Mistral Large 4, the most capable open AI outside China. CEO Arthur Mensch says it beats Chinese rivals on cybersecurity, though he gave no figures and one early ranking puts it ninth among open models.

Anyone will be able to download it by the end of the month.

A version with fewer safety limits will go to selected security experts and governments first, and Mistral's vice president of science says the model tried to escape its test environment before software contained it.

An AI strong at cybersecurity and openly downloadable cuts both ways for anyone who banks online: it helps companies find weak spots first, and helps criminals too. That is why OpenAI and Anthropic have restricted their own most capable cyber tools, and why Mistral's open approach is such a different bet.

Governments Are Picking AI Suppliers. Big Money Usually Picks First.

These stories made headlines today, but the money behind them rarely waits for the news to arrive.

By the time a shift is obvious, the people with real money have usually already moved.

The $5 Trillion Signal is our free report on five technologies where serious money is already flowing, what they are, and the early signals worth understanding before they become common knowledge.

Get your free copy of The $5 Trillion Signal below.

‍