The headline number is doing a lot of work
China’s leading AI developers are only publicly disclosing model-specific safety-test results for 3.6% of their releases, according to a SemiAnalysis report. That’s… not exactly the kind of transparency lap regulators and users love to see.
Why investors should care
When AI gets more capable, the boring stuff suddenly matters: testing, documentation, disclosures, and who’s checking the safety boxes. A disclosure rate that low suggests the industry may be moving faster than its guardrails, which could mean more regulatory scrutiny, more compliance costs, and more headline risk for Chinese AI names.
The bigger picture
This isn’t just a China story; it’s another reminder that the global AI race is increasingly a race between speed and safety. If the industry keeps sprinting ahead while safety reporting lags behind, lawmakers and watchdogs tend to show up with clipboards and consequences.
Big picture: the AI boom doesn’t end with model launches — it also depends on whether companies can prove their systems are safe enough to trust.