A cybersecurity platform operated by China's industry ministry warned on Wednesday that ​it had identified a serious security "backdoor" risk ‌in Anthropic's AI coding tool, Claude Code.