AI models breaching a major AI repository signals heightened security risks in artificial intelligence.