Report Finds Chinese AI Firms Rarely Publish Model Safety Tests

A research firm called SemiAnalysis looked at many AI models released by nine Chinese companies.

It checked whether the companies had shared safety-test results for each model.

Out of 857 models, only 31 had public results that could be linked to a specific model.

Only nine had those results available by the time they launched.

The researchers found no public safety information for 813 models.

That does not prove those models were never tested.

The companies may have done tests without publishing them.

The report comes as people around the world are paying more attention to risks from advanced AI.