The AI That Found Thousands of Zero-Days: Why Anthropic Won't Let Anyone Else Use It
Anthropic's new Mythos model can identify and exploit zero-days in every major OS and browser, and has already found thousands of critical vulnerabilities. Rather than release it publicly, the company is distributing it to 52 partner organisations through Project Glasswing. It is the first AI model withheld not for being insufficient, but for working too well.
Read full story →Alibaba Runs 10,000 of Its Own AI Chips in a Production Data Centre
Alibaba and China Telecom have launched a southern China data centre running 10,000 Zhenwu 810E chips from Alibaba's T-Head division. The deployment moves China's domestic AI chip strategy from announcement to production, and raises real questions about how much the US export control approach is actually slowing China down.
Read full story →Who Gets the Dangerous AI First Tells You Everything About Who the AI Transition Is For
Anthropic's Project Glasswing gives a model capable of compromising every major OS to Amazon, Apple, Microsoft, and 40 other organisations. The "defensive use" framing is honest, but the partner list is not neutral. It tells us exactly who the AI transition is being structured to benefit, and it is not the people most at risk from the systems being built.
Read opinion →