The US Department of Homeland Security said in July 2026 that Chinese AI company DeepSeek's models may contain malware or other security vulnerabilities