Hello, dear readers!

Here’s your weekly AI download, trimmed down so you don’t have to scroll through 47 hot takes on LinkedIn.

Today's focus — Japan and China offer alternatives to Mythos. Just two weeks after a US export ban on Anthropic, companies in Japan and China showed up with their own frontier models, ready to fill the empty shelf. Turns out, cutting people off from AI might just motivate them to build their own. Who would have thunk?

Also in this week's edition:

It's Not Just Mythos Now

The U.S. government's decision to block exports of Anthropic's most advanced AI models is already doing something predictable: creating demand for alternatives.

This week, Tokyo-based startup Sakana AI introduced Fugu, a frontier model built for AI agents and orchestration. The pitch is simple: similar performance to Anthropic's Fable 5 and Mythos Preview, but without the small inconvenience of export restrictions. Their website even leans into it with "frontier capability without the risk of export controls," which is about as subtle as it sounds.

China's cybersecurity company 360 followed with Tulongfeng, an AI model for vulnerability discovery that it says can go head-to-head with Mythos, plus Yitianzhen, focused on automated cyber defence. Founder Zhou Hongyi framed advanced vulnerability-finding AI as a strategic national asset and warned against "one-way transparency" — essentially, a world where only some countries get the good tools.

Neither company is officially calling this a reaction to Washington. Sakana says Fugu has been in development for a while, and its founders still insist U.S. models matter for Asia. But they also admit the obvious: relying on a single foreign AI provider suddenly looks like a bad idea.

The takeaway isn’t complicated. Export controls can slow things down, but they also give everyone else a reason to catch up faster. Whether these new models actually match U.S. systems is still an open question. But the race is no longer a domestic one.

OpenAI Is Building Its Google Moment

OpenAI showed up at the Cannes Lions advertising festival this week, which is usually where companies go when they’ve decided ads are no longer optional.

Speaking to the Financial Times, OpenAI's global advertising lead Dave Dugan said the company is "totally committed to ads." About one in five ChatGPT queries already has clear commercial intent, which makes it a pretty attractive place to insert a sponsored suggestion or two. For now, ads sit below responses, are clearly labelled, and — according to OpenAI — don’t influence what the model says. Make of that what you will.

The playbook is familiar. Google turned search intent into one of the most profitable businesses ever built. OpenAI is betting conversational intent can do the same — and help pay for models that keep getting more expensive to run.

Ford Learned AI Can't Replace Gray Beards

One of the more grounded AI stories this week came from Ford, which turns out still builds physical things.

According to Bloomberg, the company rehired around 350 veteran engineers after its AI-driven quality systems didn’t quite deliver the magic. The assumption had been that feeding requirements into AI would naturally produce better designs. Instead, it mostly produced a reminder that experience still matters.

Ford isn’t ditching AI. It’s just pairing it with people who know what they’re doing. Those returning engineers are now helping train younger staff and improve the AI systems themselves — a slightly less futuristic, but more effective setup.

The results are already showing up, claims the company: warranty and recall costs are down by hundreds of millions, and Ford recently topped J.D. Power's Initial Quality Study among mainstream brands. Turns out, "human in the loop" wasn’t just a buzzword after all.

Thanks for reading AIport.

Until next Monday — by then, AI will almost certainly have found another way to surprise us.