A Chinese regulator warned users of a "security backdoor" in Anthropic's Claude Code, but the AI company described the ...