BEIJING, Sept. 19 (Xinhua) -- Google's Gemini hacked into other companies during a cybersecurity test, The Wall Street Journal reported Friday, raising fresh concerns about security risks posed by U.S. AI systems.

This is not the first cybersecurity incident caused by an AI model. From an autonomous OpenAI agent escaping its sandbox to breach Hugging Face, to hackers using Anthropic's Claude to crack OpenAI's own defenses, a spate of recent mishaps among Silicon Valley's titans have spurred deep concerns that America's AI race is running dangerously off the rails.

In fact, AI security risks have already spilled over from cyberspace into the real world. A false intelligence report generated with the help of AI almost led the U.S. military to intercept and board a Chinese ship in the Middle East during the Iran war, according to an exclusive report from CNN.

Faced with a string of security incidents, some industry leaders, including Anthropic CEO Dario Amodei, have called for slowing the pace of AI development. Yet Silicon Valley's tech giants find themselves caught between safety and competition. They have preached cooperation and safety from the pulpit while quietly scrapping pledges to withhold models without proper risk mitigations, fearing that rivals could "blaze ahead," and lobbying Washington to delay AI regulation.

Their machinations have already been laid bare. Mistral, Europe's leading AI company, and others have accused their U.S. rivals of using safety concerns to entrench their market dominance and crowd out competitors.

AI may well need a pit stop, but it is the reckless front-runner who repeatedly bounces off the walls that needs to hit the brakes, not the rest of the world.

In this light, calling for a "global timeout" is neither fair nor realistic. The United States sits comfortably at the apex of computing power and algorithms, having already reaped enormous technological gains, while most other countries are only beginning to catch up.

Sound global AI governance should not be driven by Silicon Valley's fearmongering, much less by its attempts to dictate terms to the rest of the world. It requires the leading, highest-risk driver to buckle up and slow down first -- allowing the rest of humanity to share in the benefits of AI in a safe, orderly world.