TINOS, Greece, Sept 7 (Reuters Breakingviews) – Artificial intelligence may already be running out of control. Managing the risks requires an international regulatory regime. To get there, the United States will have to first accept the dangers and then cut a deal with China. AI has huge advantages. But the latest models are capable of sophisticated cyberattacks and could also help produce new deadly pathogens if they get into the hands ?of militant or criminal groups. Autonomous AI agents may increasingly go on the rampage themselves, as some OpenAI ones did in July. Aggressive countries may also use them to…