THE 60-SECOND VERSION
● NSA, CISA and FBI allege industrial-scale distillation by Chinese AI firms. Advisory, not a ban.
● Apple: $1,999 foldable, and the argument that the iPhone is the best device for AI.
● 4 days to Claude Code weekly limits settling 17 percent lower.
Three agencies warn on distillation
The NSA, CISA and FBI issued a joint advisory alleging that Chinese AI companies including DeepSeek are conducting industrial-scale distillation campaigns — systematically training on the outputs of Western models.
Distillation allegations have circulated since early 2026. What changed is who is making them: an accusation from a competitor is a commercial claim, while a joint advisory from three agencies is something a procurement team has to write down.
It is not a ban and nothing requires you to stop using Kimi K3, GLM-5.3 or Qwen. But for teams that chose open weights specifically to reduce single-vendor dependency, it reintroduces geopolitical exposure from the other direction.
What it does and does not do →
Apple sells the device instead of the model
John Ternus opened his first event as chief executive by arguing the iPhone is the best device for AI, then unveiled the iPhone Duo foldable from $1,999, an iPhone 18 Pro at $100 above last year on the A20 Pro chip, Apple Watch Series 12 and AirPods 5.
It is a coherent reframing of a real weakness. Apple has no frontier model and was never going to announce one at a hardware event nine days into a new chief executive. The hole in the argument is that the most capable models do not run on a phone, and the partner you would reach for instead has a hearing against Apple on 1 October.
The AI features, and the pricing strategy →
Also this week
- Meta launched Muse, a personal AI agent powered by Muse Spark, running on a secure VM with a separate Sentinel agent overseeing its actions. Booking travel and sending email are the named use cases.
- An Anthropic researcher resigned publicly, saying he believes neither Anthropic nor OpenAI is acting responsibly and warning about the pace toward self-improving systems. He had spent three years in pretraining research at both companies.
- ChatGPT Images got faster, with sharper detail and more refined edits.
- New research on pretraining suggests that between 2019 and 2025, roughly three times more compute-efficiency gain came from data improvements than from model architecture — with small models benefiting most from data quality.
The rest of the month
| Date | What happens |
|---|
| Sept 14 | Claude Code weekly limits settle 17 percent below current levels |
| Sept 29 | OpenAI DevDay, San Francisco |
| Oct 1 | OpenAI vs Apple hearing |
| Oct 24 | deepseek-chat and deepseek-reasoner deprecated |
| Nov 12 | OpenAI models leave Cursor |
Still open
FAQ
What did the NSA, CISA and FBI warn about?
A joint advisory alleging Chinese AI companies including DeepSeek are conducting industrial-scale distillation campaigns against Western models. It is guidance rather than a prohibition.
Should I stop using Chinese open-weight models?
Nothing requires it. Whether your organisation should is a procurement question, and it will be answered differently by a startup than by a government contractor.
How much is the foldable iPhone?
The iPhone Duo starts at $1,999.
What is the next confirmed AI deadline?
14 September, when Claude Code weekly limits settle at 25 percent above the pre-May baseline — 17 percent below current levels.