U.S. ban hits Anthropic’s top models

Anthropic just found the hard edge of AI policy

A Friday letter from Commerce cut off Anthropic’s best models worldwide. The Trump administration placed Anthropic’s Fable 5 and Mythos 5 models under export restrictions, according to WSJ. That bars use by foreign governments, companies, individuals, and foreign nationals inside the U.S.

Everyone got cut off.

Anthropic appears to have shut down broader access because sorting eligible users from banned ones is messy, risky, and legally expensive. My read is that this is the clearest sign yet that frontier AI is being treated less like software and more like controlled strategic equipment.

That matters. The immediate pain is customer access, especially for companies outside the U.S. building on these models. The bigger signal is that AI companies may now need compliance systems as serious as their research labs.

The open question is how many buyers will keep building on models that can disappear after one government letter.

Apple’s AI upgrade comes with a cutoff

Siri AI will skip plenty of devices that still feel new.

Apple’s overhauled Siri is coming this fall with iOS 27, macOS Golden Gate, watchOS 27, and other updates, but only newer hardware will support it. The list starts with iPhone 15 Pro, M1 iPads, Apple silicon Macs, Apple Watch Series 9, and all Vision Pro models.

That cutoff is the story. My read is that Apple is turning AI into a hardware filter, which makes technical sense and also helps sell upgrades. The sharper detail is that some Siri voice controls and Apple’s most powerful on-device model need even newer machines with at least 12GB of unified memory.

This will be annoying for people holding older iPhones in good condition at the kitchen table.

The risk for Apple is that Siri AI feels like a paid hardware feature before it has proved itself. The bigger question is whether users upgrade because Siri is useful, or because Apple quietly made older devices feel outside the future.

India’s AI advantage may be cost

A five-second 720p video clip now costs less than a rupee to generate.

Avataar AI’s new Varya model is built for Indian e-commerce and local context, including festivals, food, clothing, and architecture. The company says it can create that clip in 45 seconds on an Nvidia H200, compared with 1,230 seconds for Alibaba’s Wan 2.2.

That speed matters. My read is that Avataar is making the right tradeoff for India’s market: cheaper, narrower, useful video beats chasing the biggest general model. The ₹0.48 per second price is the real story because most Indian businesses will not build around tools that cost $0.10 per second or more.

You can almost hear the pitch being made in a bright office: faster ads, lower costs, local faces, local clothes.

The open-weight release on AIKosh also gives India’s developer ecosystem something concrete to test and adapt. The question is whether “good enough and cheap” can become a durable advantage, or whether bigger labs copy the same playbook fast.

Prohuman team

Free Guides

Explore our free guides and products to get into AI and master it.

All of them are free to access and would stay free for you.

Feeling generous?

You know someone who loves breakthroughs as much as you do.

Share The Prohuman it’s how smart people stay one update ahead.