Only 31 of China's AI model releases came with a safety report.
SemiAnalysis, a chip research firm, published a count on Thursday of 857 model releases from nine Chinese developers. Only 31 ever came with a published safety evaluation, and nine had one by launch day. The authors stress that a missing report means none was found, not that nobody tested.
Why it mattersIf you run an open Chinese model, you are probably trusting it without any published safety check. Zhipu is the exception, with a result every year since 2022.
Read in the app — free, in 9 languages
Related stories
Mathematicians asked their field to stop working with OpenAI.2026-10-10A new scoreboard grades how often agents do what you never asked.
2026-10-09Google's medical AI interviewed patients before their doctor did.
2026-10-09Firms funding a $1.8 billion biology dataset get to use it first.
2026-10-09A sign error wiped out three of OpenAI's 722 new maths papers.
2026-10-08