The Sui blockchain architect is leaving crypto infrastructure for defensive AI research, and his reasons explain a lot about where the industry is headed
The Claude maker is reportedly assembling an in-house silicon team ahead of a 2026 target, joining a growing list of AI companies cutting out the middleman
The UK's AI Safety Institute just published the most serious autonomous misbehavior incident on record — and the implications go well beyond the two models involved
Europe's first independent evaluation of Zhiyu AI's open-weight model finds capability gaps of two to four months behind the frontier, but safety gaps that are far wider