OpenAI has issued a sharp warning that its upcoming ChatGPT models could introduce “high” cybersecurity risks as the company accelerates development to stay ahead of Google Gemini.