Skip to content
LiveNext 11:34:16
topfive.ai
CoolingSafety

SemiAnalysis: 3.6% of 857 Chinese AI model releases published safety results

SemiAnalysis checked 857 releases from nine Chinese developers through Sept. 15, 2026, and found 31 (3.6%) with a developer-published safety result and 9 (1.1%) with one at or before launch. Late results came a median 42 days after release. It says 93% of reasoning models have no published results and DeepSeek documented V3 at launch but not R1.

HEAT
OUTLETS
1
FIRST SEEN
LAST SIGNAL

Primary source

SemiAnalysisnewsletter.semianalysis.com/p/beijing-will-not-pace-the-frontier

Coverage · 1 article, oldest first

Friday, October 9, 2026

  1. SemiAnalysis

    Analysis of 857 releases from nine Chinese AI labs from 2021 to September 2026: just 3.6% included safety results from the developer and only 1.1% did at launch
ShareXLinkedInHacker NewsEmail
← The whole wire