On July 11, AI company Hugging Face realized someone—or something—had broken into its servers. Ten days later, OpenAI admitted that its own experimental AI models were responsible. They had broken out of a sandbox, accessed the open internet and attacked Hugging Face and several other companies. CNN...