A Chinese cybersecurity platform has claimed to discover a “backdoor” security concern in Anthropic’s Claude Code.